Contents | Zoom in | Zoom out Search Issue | Next Page

Transcription

Contents | Zoom in | Zoom out
For navigation instructions please click here
Search Issue | Next Page
Contents | Zoom in | Zoom out
For navigation instructions please click here
Search Issue | Next Page
Previous Page | Contents | Zoom in | Zoom out | Front Cover | Search Issue | Next Page
M
q
M
q
M
q
M
q
MQmags
q
THE WORLD’S NEWSSTAND®
CALL FOR PAPERS
IEEE Geoscience and Remote Sensing Magazine
This is the third issue of the new IEEE Geoscience and Remote Sensing Magazine, which was approved by the IEEE
Technical Activities Board in 2012. This is an important achievement for GRSS since it has never had a publication in
the magazine format. The magazine will provide a new venue to publish high quality technical articles that by their very
nature do not find a home in journals requiring scientific innovation but that provide relevant information to scientists,
engineers, end-users, and students who interact in different ways with the geoscience and remote sensing disciplines.
The magazine will publish tutorial papers and technical papers on geoscience and remote sensing topics, as well as
papers that describe relevant applications of and projects based on topics addressed by our society.
The magazine will also publish columns on:
- New satellite missions
- Standard remote sensing data sets
- Education in remote sensing
- Women in geoscience and remote sensing
- Industrial profiles
- University profiles
- GRSS Technical Committee activities
- GRSS Chapter activities
- Conferences and workshops.
The new magazine is published in with an appealing layout, and its articles will be included with an electronic format
in the IEEE Xplore online archive. The magazine content is freely available to GRSS members.
This call for papers is to encourage all readers to prepare and submit articles and technical content for review to be
published in the IEEE Geoscience and Remote Sensing Magazine. Contributions for the above-mentioned columns of
the magazine are also welcome.
All technical papers will undergo blind review by multiple reviewers. The submission and the review process is managed
at the IEEE Manuscript Central as it is already done for the three GRSS journals. Prospective authors are required to
submit electronically using the website http://mc.manuscriptcentral.com/grs and selecting the “Geoscience and Remote
Sensing Magazine” option from the drop-down list. Instructions for creating new user accounts, if necessary, are available on the login screen. No other manners of submission are accepted. Papers should be submitted in single column,
double-spaced format. The review process will assess the technical quality and/or the tutorial value of the contributions.
The magazine will publish also special issues. Readers interested to propose a special issue can contact the Editor In Chief.
For any additional information and for submitting papers contact the Editor In Chief:
Prof. Lorenzo Bruzzone
University of Trento,
Trento, Italy
E-Mail: ______________________
[email protected]
Phone: +39 0461 28 2056
Digital Object Identifier 10.1109/MGRS.2013.2277688
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
SEPTEMBER 2013
VOLUME 1, NUMBER 3
WWW.GRSS-IEEE.ORG
_____________
FEATURE
6
A Tutorial on Speckle Reduction
in Synthetic Aperture Radar Images
© WIKIMEDIA COMMONS/OLDTECHED
by Fabrizio Argenti, Alessandro Lapini,
Luciano Alparone, and Tiziano Bianchi
PG. 6
SCOPE
ON THE COVER:
Synthetic Aperture Radar is a very
important technology for Earth
Obs e r v at ion . F or i mprov i ng
understanding of SAR images it is
important to use data processing
techniques for speckle reduction.
© WIKIMEDIA COMMONS/ASTRIUM GMBH
IEEE Geoscience and Remote Sensing Magazine will inform readers of
activities in the GRS Society, its technical committees, and chapters.
GRSM will also inform and educate readers via technical papers, provide
information on international remote sensing activities and new satellite
missions, publish contributions on education activities, industrial and
university profiles, conference news, book reviews, and a calendar of
important events.
SEPTEMBER 2013
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE
1
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
COLUMNS &
DEPARTMENTS
3 FROM THE EDITOR
5 PRESIDENT’S MESSAGE
36 TECHNICAL COMMITTEES
39 CHAPTERS
42 EDUCATION
46 WOMEN IN GRS
48 CONFERENCE REPORTS
61 GRSS MEMBER HIGHLIGHTS
63 CALENDAR
EDITORIAL BOARD 2013
Dr. Lorenzo Bruzzone
Editor-in-Chief
University of Trento
Department of Information Engineering
and Computer Science
Via Sommarive, 5
I-38123 Povo, Trento, ITALY
E-mail: ______________
[email protected]
Dr. William Blackwell
MIT Lincoln Laboratory
Lexington, MA 02420-9108, USA
E-mail: [email protected]
_______
Dr. Kun Shan Chen
National Central University
Chungli, TAIWAN
____________
Dr. Paul Gader
CISE Dept., University of Florida
301 CSE Bldg.
Gainesville, FL 32611, USA
_________
Dr. John Kerekes
Rochester Institute of Technology
54 Lomb Memorial Dr.
Rochester, NY 14623, USA
_________
Dr. Antonio J. Plaza
Department of Technology of Computers
and Communications
Escuela Politecnica de Caceres,
University of Extremadura
Avda. de la Universidad S/N
E-10071 Cáceres, SPAIN
________
Dr. Gail Skofronick Jackson
NASA Goddard Space Flight Center
Code 612
Greenbelt, MD 20771, USA
____________
Dr. Stephen Volz
NASA Earth Science Div.
300 E St., SW
Washington, DC 20546, USA
E-mail: ________
[email protected]
MISSION STATEMENT
The IEEE Geoscience and Remote Sensing Society of the IEEE seeks to advance science and
technology in geoscience, remote sensing and
related fields using conferences, education, and
other resources.
IEEE Geoscience and Remote Sensing Magazine (ISSN 2168-6831) is published
quarterly by The Institute of Electrical and Electronics Engineers, Inc., IEEE
Headquarters: 3 Park Ave., 17th Floor, New York, NY 10016-5997, +1 212 419
7900. Responsibility for the contents rests upon the authors and not upon
the IEEE, the Society, or its members. IEEE Service Center (for orders, subscriptions, address changes): 445 Hoes Lane, Piscataway, NJ 08854, +1 732
981 0060. Price/Publication Information. Subscriptions: included in Society
fee for each member of the IEEE Geoscience and Remote Sensing Society.
Nonmember subscription prices available on request. Copyright and Reprint
Permissions: Abstracting is permitted with credit to the source. Libraries are
permitted to photocopy beyond the limits of U.S. Copyright Law for private
use of patrons: 1) those post-1977 articles that carry a code at the bottom of
GRS OFFICERS
President
Dr. Melba M. Crawford
Purdue University, USA
Executive Vice-President
Dr. Kamal Sarabandi
University of Michigan, USA
Vice-President of Meetings and Symposia
Dr. Adriano Camps
Universitat Politecnica de
Catalunya-Barcelona Tech, Spain
Vice-President of Publications
Dr. William Emery
University of Colorado, USA
Vice-President of Technical Activities
Dr. John Kerekes
Rochester Institute of Technology, USA
Vice-President of Professional Activities
Dr. Wooil M. Moon
University of Manitoba, Canada
Vice-President of Information Resources
Dr. Steven C. Reising
Colorado State University, USA
IEEE PERIODICALS
MAGAZINES DEPARTMENT
Associate Editor
Laura Ambrosio
Senior Art Director
Janet Dudar
Assistant Art Director
Gail A. Schnitzer
Production Coordinator
Theresa L. Smith
Business Development Manager
Susan Schneiderman
+1 732 562 3946
__________
[email protected]
Fax: +1 732 981 1855
Advertising Production Manager
Felicia Spagnoli
Production Director
Peter M. Tuohy
Editorial Director
Dawn Melley
Staff Director, Publishing Operations
Fran Zappulla
the first page, provided the per-copy fee indicated in the code is paid through
the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923 USA;
2) pre-1978 articles without fee. For all other copying, reprint, or republication
information, write to: Copyrights and Permission Department, IEEE Publishing
Services, 445 Hoes Lane, Piscataway, NJ 08854 USA. Copyright © 2013 by the
Institute of Electrical and Electronics Engineers, Inc. All rights reserved. Postmaster: Send address changes to IEEE Geoscience and Remote Sensing Magazine, IEEE,
445 Hoes Lane, Piscataway, NJ 08854 USA. Canadian GST #125634188
PRINTED IN USA
IEEE prohibits discrimination, harassment, and bullying. For more information,
visit http://www.ieee.org/web/aboutus/whatis/policies/p9-26.html.
_____
2
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
FROM THE EDITOR
BY LORENZO BRUZZONE
T
his third issue of the new IEEE Geoscience and Remote
Sensing Magazine has been published shortly after
the recent premier international remote sensing conference, IGARSS 2013 in Melbourne, Australia.
One important news item regarding the magazine is that the IEEE Geoscience and Remote Sensing
Administrative Committee decided during their July
2013 meeting to distribute both digital and electronic
versions of the magazine to all GRSS members free
of charge during 2013 and 2014. This will allow all
GRSS members to access the magazine’s contents
and to become familiar with the digital format of the
magazine. The digital format is different from the
electronic format, which is used by the IEEE Xplore
web site. The digital format allows readers to navigate
and explore the technical content of the magazine
with a look and feel similar to that of an issue of a
printed magazine. The electronic version provided on
the IEEE Xplore web site does not have a magazine layout and only allows readers to access each article as a
separate pdf file.
Another important news item on the magazine is
that Manuscript Central, the online submission and
review web site, is open to accept article submissions. I
would like to remind you that the goal of the magazine
is to publish tutorial and technical papers on geoscience and remote sensing topics, as well as papers that
describe relevant projects based on and applications of
topics addressed by the GRS Society. All tutorial and
technical papers will undergo blind review by multiple
reviewers. In addition, the magazine publishes regular
columns on education in remote sensing, remote sensing missions, standard data sets, women in geoscience
and remote sensing, space agency news, book reviews,
etc. Please refer to the call for papers in this issue for
Date of publication: 1 October 2013
SEPTEMBER 2013
more details on the topics of the magazine and on the
procedure for submitting your contributions to the magazine. I would like to encourage everyone to prepare and
submit articles and technical content for review to be
published in the Magazine.
This issue includes an excellent Feature article,
which is a tutorial paper on despeckling methods for
Synthetic Aperture Radar (SAR) images prepared by
well-known and highly respected scientists in the field
from Italy. The paper provides a critical review of the
main despeckling techniques
presented in the literature since
their initial use over 30 years
GRSM DIGITAL AND ELECago, highlighting trends and
TRONIC VERSIONS WILL BE
changes in the approaches to
DISTRIBUTED FREE OF
despeckling over years. Various fundamental and advanced CHARGE TO ALL GRSS MEMBERS IN 2013 AND 2014.
concepts are presented with
examples and critical discussion, thereby providing an
updated and comprehensive overview of this topic.
Moreover, future potential methods based on new
concepts of signal processing, including compressive
sensing, are discussed.
The Technical Committees column contains an article
announcing the results of the 2013 IEEE GRSS Data
Fusion Contest organized by the Data Fusion Technical
Committee. After a brief overview of previous contests,
the details of this year’s event are reported, including the
outcome of the contest.
The Women in Geoscience and Remote Sensing presents
a short article written by Prof. Florence Tupin, ParisTech,
France. She is an international leader in image processing in remote sensing and provides an example of a very
successful career.
The Chapters section contains an article that introduces the recently established GRSS Turkey Chapter.
The article briefly describes the current activities of the
3
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
new chapter, which is warmly welcome in the GRSS community! This section also reports an updated table with all
the active GRSS chapters.
The Education column includes a report on the Remote
Sensing “Summer” School 2013 held on July 18–19, 2013,
in Melbourne, Australia, just before IGARSS 2013. The article provides a summary of the activities of this very successful school (which, considering its geographical location,
was a Winter school rather than a Summer school).
The Conference Reports section contains two contributions. The first is an article describing the GRSS Major
Awards and Fellow Recognitions presented at the IGARSS
2013 Plenary Session held on July 22, 2013, in Melbourne,
Australia. The article describes the excellent organization
of IGARSS 2013 and provides information on all the Major
URSI Commission F
Microwave Signatures 2013
Specialist Symposium on Microwave Remote
Sensing of the Earth, Oceans, and Atmosphere
October 28-31, 2013
Espoo, Finland
URSI Commission F
Aalto University
Awards and Fellow recipients. Congratulations to all of
them! The second report addresses the 7th International
Workshop on the Analysis of Multi-Temporal Remote Sensing Images (MultiTemp 2013) held on June 25–27, 2013, in
Banff, Alberta, Canada. This is one of the workshop series
that have been technically cosponsored by IEEE GRSS since
their first offering and that have been a growing success
within our community.
Finally, I would like to draw your attention to the various calls for nominations and calls for papers in this issue.
I wish everyone an enjoyable and productive autumn.
Lorenzo Bruzzone
Editor-in-Chief
GRS
ICMARS 2013
9th International Conference
on Microwaves, Antenna,
Propagation & Remote Sensing
December 11 - 14, 2013
Pre Conference Workshop 10th Dec. 2013
Jodhpur, INDIA
Sponsored by IEEE GRSS
General Chair:
Prof. Martti Hallikainen, Aalto University
Abstract submission:
May 20, 2013
One-page abstract
Email: [email protected]
Notification of abstract review outcome:
July 15, 2013
Advance registration:
September 10, 2013
ICMARS 2013 Chair:
Prof. OPN Calla, Director, ICRS, Jodhpur
IMPORTANT DATES:
Abstract Submission Deadline:
Aug 30, 2013
Abstract Acceptance Notification:
Sep 05, 2013
Early Bird Registration:
Sep 25, 2013
Deadline for Authors to Register:
Oct 15, 2013
Deadline for Full length paper Submission: Nov 01, 2013
Abstract Submission deadline Likely to be extended
Web Address:
http://www.icmars2013.org
Web Address:
http://frs2013.ursi.fi
4
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
PRESIDENT’S MESSAGE
BY MELBA CRAWFORD
J
uly marks the end of an era for the GRSS Adcom,
marked by the departure of Tammy Stein, who has
served the Society for more than 20 years in multiple capacities. We wish Tammy great success in her new business venture, and we will greatly miss her professional
contributions and her personal friendship over these
many years.
On behalf of the IEEE Geoscience and Remote Sensing Society, I congratulate the IGARSS 2013 local organizing team, led by General Co-Chairs Peter Woodgate
and Simon Jones, on the outstanding IGARSS that was
held July 21–26 in Melbourne, Australia. The Symposium was attended by more than 1350 delegates from
66 countries. President Peter Staecker, the President of
IEEE, spoke at the opening session, and was joined by
a number of plenary speakers, including Professor Guo
Huadong, Director-General of the Institute of Remote
Sensing and Digital Earth of the Chinese Academy of
Sciences, Dr. Chris Pigram, Chief Executive Officer of
Geoscience Australia, Dr. Mike Goodchild, Professor
Emeritus of Geography from U. of California Santa Barbara, and Dr. Rob Vertessy, Director of the Australian
Bureau of Meteorology.
IGARSS provides a great venue to celebrate the many
contributions of our members. President Staecker presented the IEEE Electromagnetics Award to Leung Tsang
at the Plenary Session. Roger Lang received the GRSS
Distinguished Achievement Award, Kamal Sarabandi
was selected for the GRSS Education Award, and Carlos
Lopez received the GRSS GOLD award. We also recognized the 10 newly elevated IEEE Fellows evaluated by
the GRS Society. Winners of the publications awards
and the 2013 Student Prize Paper Competition were announced and presented at the banquet.
SEPTEMBER 2013
The Symposium was preceded by a very successful
Remote Sensing Summer School, organized by Xiuping
Xia, Kim Lowell, and Mike Inggs, the GRSS Education
Director, which was attended by more than 50 graduate students. The IGARSS 2013 Technical Program was
quite successful, with 12 parallel oral sessions, followed
by the evening poster sessions. Our attendees appreciated the extended schedule for the exhibits which provided opportunities to engage, and we are grateful to the
sponsors for their support.
The social activities are always a highlight for
IGARSS, and 2013 did not disappoint. The opening
reception launched the Symposium with outstanding
food and piano accompaniment. The Young Professionals Luncheon was a sellout activity—and Peter Staecker
challenged attendees to engage and to make a difference, but to also consider work-life balance. The Women
in Geosciences reception became a great opportunity for
networking over wine and hors d’oeuvres, and everyone
enjoyed the reception in the midst of the saltwater tanks
at the Melbourne Aquarium. The symposium banquet
was a true epicurean delight in the Plaza Ballroom.
Attendees reluctantly left Melbourne after a terrific
week—some off to explore more of Australia and the
rest back to our real lives after an amazing experience!
We look forward to seeing you in Quebec City, Canada,
next year. Although IGARSS moves from one country
to another every year, its permanent home is the IEEE
Geoscience and Remote Sensing Society. Until next July,
please join us at chapter meetings, join a technical committee, and contribute to our publications.
Best Wishes,
Melba Crawford
President, IEEE GRSS
[email protected]
______________
GRS
5
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
FABRIZIO ARGENTI, ALESSANDRO LAPINI,
AND LUCIANO ALPARONE
Department of Information Engineering, University of Florence,
Via S. Marta, 3, 50139, Florence, Italy
TIZIANO BIANCHI
Department of Electronics and Telecommunications, Politecnico di Torino,
Corso Duca degli Abruzzi, 24, 10129, Torino, Italy
Abstract—Speckle is a granular disturbance, usually modeled as a multiplicative noise, that affects synthetic
aperture radar (SAR) images, as well as all coherent images. Over the last three decades, several methods have been
proposed for the reduction of speckle, or despeckling, in SAR images. Goal of this paper is making a comprehensive
review of despeckling methods since their birth, over thirty years ago, highlighting trends and changing approaches
over years. The concept of fully developed speckle is explained. Drawbacks of homomorphic filtering are pointed
out. Assets of multiresolution despeckling, as opposite to spatial-domain despeckling, are highlighted. Also advantages of undecimated, or stationary, wavelet transforms over decimated ones are discussed. Bayesian estimators and
probability density function (pdf) models in both spatial and multiresolution domains are reviewed. Scale-space
varying pdf models, as opposite to scale varying models, are promoted. Promising methods following non-Bayesian
approaches, like nonlocal (NL) filtering and total variation (TV) regularization, are reviewed and compared to spatial- and wavelet-domain Bayesian filters. Both established and new trends for assessment of despeckling are presented. A few experiments on simulated data and real COSMO-SkyMed SAR images highlight, on one side the costperformance tradeoff of the different methods, on the other side the effectiveness of solutions purposely designed
for SAR heterogeneity and not fully developed speckle. Eventually, upcoming methods based on new concepts of
signal processing, like compressive sensing, are foreseen as a new generation of despeckling, after spatial-domain
and multiresolution-domain methods.
I. INTRODUCTION
ynthetic aperture radar (SAR) remote sensing [1] offers a number of advantages over optical remote sensing, mainly the all-day, all-weather acquisition capability.
However, the main drawback of SAR images is the presence
of speckle, a signal dependent granular noise, inherent of
all active coherent imaging systems, that visually degrades
the appearance of images. Speckle may severely diminish
the performances of automated scene analysis and information extraction techniques, as well as it may be harmful in applications requiring multiple SAR observations,
like automatic multitemporal change detection. For these
reasons, a preliminary processing of real-valued detected
SAR images aimed at speckle reduction, or despeckling, is
This work has been carried out under the financial support of Italian Space
Agency (ASI), COSMO-Skymed AO ID #2293.
S
6
2168-6831/13/$31.00©2013IEEE
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
© WIKIMEDIA COMMONS/OLDTECHED
of crucial importance for a number of applications. Such
a preprocessing, however, should be carefully designed to
avoid spoiling useful information, such as local mean of
backscatter, point targets, linear features and textures.
A steadily increasing number of papers specific on
despeckling has appeared in the literature over the last
ten years, presumably because the new generation of satellite SAR systems has dramatically raised the attention
of researchers in signal processing towards this problem. The COSMO-SkyMed constellation—four satellites
launched by the Italian Space Agency (ASI) between 2007
and 2010—features X-band SAR with low revisit-time; as
a second generation mission, two additional satellites are
foreseen in 2014 and 2015. The twin-satellite constellation
TerraSAR-X/TanDem-X (2007/2010) launched by the German Space Agency (DLR) and the upcoming Sentinel-1a/1b satellite constellation (2013/2015) from the European
Space Agency (ESA), which shall extend the EnviSat misSEPTEMBER 2013
sion, complete the European scenario of satellite SAR. Also,
the Canadian RADARSAT 3 mission is expected in a near
future, with 3 satellites operating at C-band, to be launched
in 2017. A thorough overview of past, present and future
missions can be found in [1].
The most recent advances in despeckling pursue the
technological objective of giving an extra value to the huge
amount of data that are routinely collected by current and
upcoming SAR systems mounted on orbiting platforms. In
fact, with the exception of applications related to production of digital elevation models (DEMs) or interferometric
phase maps useful for studies of terrain deformation (landslides, subsidence, etc.), SAR data do not find the same full
utilization, as optical data do, by either users’ or scientists’
communities. As an example, the functional development
of efficient techniques for fusion between optical and SAR
data would constitute an enabling technology that would
allow a relevant number of new applications to bring
7
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
benefits both for data providers and for producers of software applications. Unfortunately, speckle is the main
obstacle towards the development of an effective opticalSAR fusion [2], together with the different acquisition
geometry of optical and SAR systems.
This article is intended as tutorial on despeckling, rather
than a simple review of despeckling methods. Therefore,
emphasis is given to speckle and reflectivity models that are
used for filtering. The review of methods that are not relying
on the multiplicative noise model is kept very concise, since
such methods have not encountered same progresses over
time as model-based methods have. With only very few
exceptions, despeckling methods specifically proposed for
ultrasound images are not reviewed here. In fact, notwithstanding the similarity of the coherent imaging system, in
the presence of weak echoes from tissues the additive white
Gaussian noise (AWGN) introduced by the electronics
cannot be neglected and the noise model also includes an
AWGN term [3]. In SAR images, the AWGN term is always
negligible, compared to the signal-dependent term [4].
Despeckling methods specifically pertaining to polarimetric SAR are not discussed in this tutorial. Readers interested to this topic are addressed to the seminal papers by
Novak and Burl [5], Lee et al. [6] and Touzi and Lopès [7], as
well as to more recent and developed contributions [8], [9].
Also despeckling methods designed to take advantage of
the availability of a temporal sequence of SAR images, like
[10], [11], are not addressed here. By default, speckle reduction is approached as mono-variate, even though a multivariate speckle reduction [12], if applicable, may be preferred.
Only incoherent processing, that is, processing of SAR
images in either power, referred to as intensity, or amplitude
formats is dealt with. In fact, coherent processing of data in
complex format does not increase the signal-to-noise ratio
(SNR), but only the coherence [13] of the interferogram and
thus it is used in SAR interferometry [14]. The noisy phase
of the complex interferogram may be filtered before it is
unwrapped [15], [16], but in many cases the regularization
is directly performed by the unwrapping algorithm [17].
Whenever two real valued detected SAR images of the same
scene are available, the temporal correlation of speckle conveys information on the coherence of the interferogram
Aie jzi
A1e jz1
A2e jz2
FIGURE 1. Scattering model explaining fully developed speckle.
8
that would be calculated from complex data [18]. Thus,
speckle is not only noise but in some sense has an information content, even if difficult to exploit.
Excellent reviews of despeckling methods with high
tutorial value have been written by Lee et al. [19] and Touzi
[20]. Our goal is to update the presentation of methods to
changing times, especially towards recently established
concepts of multiresolution processing. A brief perspective
on upcoming promising approaches to despeckling in particular, and to information extraction from SAR images in
general, is also included in this survey.
The organization of contents is the following. After a leisurely paced section on fundamentals of reflectivity, speckle
and imaging system modeling, the problem is addressed
and developed under a statistical signal processing perspective, as in [20]. Emphasis is given to Bayesian estimation in
either space or scale-space domains. The main features of
the latter are concisely surveyed. A comprehensive critical
review of the most relevant speckle filters, beginning with
the pioneering Lee filter [21], spans over thirty years and
highlights trends and fashions that have been pursued and
developed over time or quickly abandoned. The renewed
interest of researchers towards despeckling occurred with
the introduction of multiresolution analysis, when spatial
domain methods had reached a degree of sophistication,
together with a saturation of performances, that demanded
a cross-fertilization from other fields of signal processing.
A variety of wavelet-based, or more generally scale-space,
despeckling methods is contextualized and discussed.
Advantages of such methods over spatial domain methods
is pointed out. Promising approaches like nonlocal filtering
and total variation regularization are described.
The second part of the article contains a more articulated
review of a few selected methods, some of them recently
proposed by the authors, that are presently indicated as
highly performing [22] in a comparative assessment carried
out on image specimens produced by a SAR simulator [23].
The most established and widely used statistical indexes to
assess the quality of despeckling, both with a reference, like
in the case of the SAR simulator, or of synthetically speckled optical images, and blind, i.e., without a reference, are
surveyed. A brief section compares quantitative results of
the selected methods and draws some considerations on
the specific features of the different methods, which exhibit
different behaviors, and on the suitability and limitations
of the quality indexes. The trade off between performances
and computational cost is analyzed. The influence of
speckle correlation on the despeckling accuracy of singlelook images and a viable strategy for its preliminary reduction, without affecting the subsequent despeckling stages,
is described. Eventually, a concluding section remarks the
key points of the analysis and gives hints that may help
researchers develop new and better despeckling filters, or
more simply may help users choose the most suitable filter
among those that are presently available, also as source or
executable codes to download.
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
II. SIGNAL AND NOISE MODELING
Under a statistical signal processing perspective, despeckling filters aim at estimating the noise-free radar reflectivity from the observed noisy SAR image [20]. In order to
describe the estimation methods that have been developed
for the despeckling problem, we need firstly to introduce
models for speckle, SAR system and reflectivity.
A. SPECKLE MODELS
SAR is an active acquisition instrument that produces a
radiation and captures the signals backscattered from a
small area of the imaged scene (resolution cell). The received
signal, as output from the in-phase and quadrature channels, is complex. If we assume that the resolution cell contains several scatterers and that no one yields a reflected
signal much stronger than the others (distributed target),
then the received signal can be viewed as the incoherent
sum of several backscattered waves, i.e., Ae jz = / i A i e jz i, as
shown in Fig. 1. The amplitudes A i and phases z i are the
result of several factors, including propagation attenuation,
scattering of the illuminated targets, antenna directivity.
Each individual component, however, can not be resolved
within a resolution cell. A first approach to modeling the
received signal is solving the Maxwell’s equations according
to the propagation geometry and scattering medium [24],
[25]. By using this approach, the way each propagation path
interferes gives us basic information about the observed
scene. On the other hand, if we consider that the phases
of each path are highly different and that they may sum
in a constructive or destructive way, then the amplitude of
the received signal varies randomly. So, even if the underlying reflectivity field is uniform, it appears as affected
by a “granular” noise after the imaging system. For visual
inspection and for specific applications that involve visual
information retrieval, such as mapping and segmentation,
the highly varying nature of the signal may be considered as
a disturbance and is commonly denoted as “speckle”.
The phases z i are highly varying (since the wavelength
is much shorter than the resolution cell size and scatterers
distances) and may be considered as uniformly distributed
in (-r, r) as well as independent of A i . If the number of
scatterers is sufficiently high, the central limit theorem
applies [26] and the resulting signal Ae jz = z 1 + jz 2 can be
seen as a complex signal whose real and imaginary parts
(in-phase and quadrature components) are independent
and identically distributed zero-mean Gaussian variables
with variance v/2. When this applies speckle is termed as
fully developed [27]. The joint probability density function
(pdf) is given by
2
2
z1 + z2
1
p z1, z2 (z 1, z 2) = rv e - v ,
(1)
whereas the amplitude A is distributed as a Rayleigh pdf,
that is
2
2A A
p A (A) = v e - v
SEPTEMBER 2013
(2)
and the power or intensity I = A 2 is distributed according to
an exponential pdf, that is
1 I
p I (I) = v e - v
(3)
so that the mean of the intensity is equal to v. It can be
shown [4], [28] that the intensity measurement carries
information about the average backscattering coefficient
(for distributed targets) related to the resolution cell.
Hence, for specific applications, the parameter v is the
actual information we would like to extract from a single
channel SAR system. This can be considered as the radar
cross section (RCS) of the observed resolution cell. The
received signal pdf can be reformulated into
1 I
p I v (I v) = v e - v
(4)
I = vu ,
(5)
or
where u is exponentially distributed, that is,
p u (u) = e -u .
(6)
Eq. (5) is termed the multiplicative model of speckle.
If only one image (realization of the stochastic process)
is available, the best estimate of the scene average reflectivity is just the pixel-by-pixel intensity. This will be a quite
noisy estimate because of the previously described constructive/destructive combination effects. From (3), it follows that the variance of the intensity in each pixel is v 2,
so that brighter pixels will be affected by stronger disturbances than darker ones. A way to improve the estimation
of v is to average L independent intensity values related to
the same position. This processing, named “multilooking”,
maintains the mean intensity v but reduces the estimator
variance to v 2 /L. Independent “looks” of a target resolution cell can be obtained either by appropriate processing
in the Doppler domain (splitting the Doppler bandwidth
within the imaging system that compensates the quadratic
phase variation created by the platform movement) or by
averaging L spatial observations. In both cases, the cost to
be paid for estimation accuracy improvement is spatial resolution loss by a factor L. If the hypothesis of independent
intensity measurements holds (in the case of correlated
data the assumption fails), the L-look averaged intensity
I L is C-distributed, that is
p IL v (I L v) =
1 ` L jL L - 1 - LIvL
I
e
C (L) v L
(7)
whereas the relative amplitude image A L = I L has a square
root C distribution [4]. For visual inspection and for automatic interpretation tasks, the use of amplitude images is
preferable, thanks to their reduced dynamic range with
respect to intensity images, which is accompanied by an
increment in SNR.
9
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
The model described above is valid under the assumption that the imaged scene is characterized by distributed
scatterers. In the presence of a scatterer much stronger than
the others (point target), the received signal pdf becomes a
Rice distribution and the model above described does not
apply. In this case, the received signal power is related to
the single target reflection coefficient and, for the purpose
of speckle removal, point targets are treated separately
from distributed targets.
B. SAR SYSTEM MODEL
In the above analysis, the effect of the imaging system has
not been taken into consideration. Indeed, the SAR system
can achieve a spatial resolution of the order of the antenna
size only if proper processing, referred to as focusing, is
applied. The energy of the transmitted frequency modulated (FM) chirp pulse is spread into the range-Doppler
domain and such a processing consists of matched filtering
along the range and along iso-Doppler curves and is needed
to compact energy back in the spatial domain [28]. From
this point of view, a SAR system can be seen as an encoding
transfer function h e (r) followed by a compression transfer
function h c (r) [4], [29]. If S (r) denotes the complex reflectivity, the observed single look complex (SLC) signal after
the imaging processor is
g c (r) = [C $ S (r) ) h e (r) + n (r)] ) h c (r),
(8)
where the constant C absorbs propagation information
(e.g., loss and antenna gains) and the term n (r) accounts
for thermal noise at the receiver. For sufficiently high signal-to-noise ratios, the noise term can be neglected and the
received complex signal becomes
g c (r) = C $ S (r) ) h e (r) ) h c (r) = C $ S (r) ) h (r).
(9)
For well-designed SAR, the impulse response h (r) is pulselike and represents the point spread function (PSF) of the
system that, in a first approximation, can be assumed as
independent of the position. Again, the intensity g c (r) 2
is proportional to the average backscattering coefficient
of the cell and is the information we would like to achieve
from the observation. An accurate description of the model
in (9) and of the statistical properties of the acquired SAR
image is given in [29].
C. REFLECTIVITY MODELS
The speckle formation model yields a pixelwise description of the observed signal. For many applications,
including despeckling, more refined models are needed.
Such models describe the observed received signal at a
coarser scale than the single pixel one and try to intercept
information about the underlying texture of the imaged
scene and its correlation. It is then crucial to consider also
the average intensity, i.e., RCS v, which is considered the
information to be retrieved, as a random process. Unfor10
tunately, the RCS is not directly observable and its properties must be inferred from the intensity values over an
area in which the texture is homogeneous. In this sense,
RCS modeling can be seen as an inverse problem whose
solution is made difficult by the fact that homogeneity
can be stated only if a ground truth is available, but often
this is not the case. Furthermore, since the problem can
be formulated only in a statistical sense, the dimension
of the homogeneous area becomes crucial: it should be
as large as possible in order to reliably apply statistical
hypothesis testing methods, but this contrasts with the
natural scenes structure that is often characterized by
the presence of limited size homogeneous areas (such as
fields, woods, orchards, forests, trees, man-made areas)
and mixing the information of different textures makes
the hypothesis tests to fail.
The starting point for solving this inverse problem is
the statistics of the observed intensity over a homogeneous
area. The pdf of the intensity signal can be written as
p (I) =
# p (I
v) p (v) dv ,
(10)
where p (I v) is the single pixel speckle model, given by
(4) and (7) for the 1-look and L-look cases, respectively. Eq.
(10) is referred to as the product model of the observed intensity [20]. One of the assumptions that must be made to state
the validity of the model (10) is that the RCS fluctuation
scale is larger than that of speckle.
Even though several pdfs have been proposed for the
intensity I (e.g., Weibull, log–normal), one of the most used
pdf is the K distribution. The K distribution is a parametric pdf that, with a suitable choice of its parameters, well
fits observed intensity histograms. It has also the advantage
that a closed form of the RCS pdf, i.e., p (v), exists such that
the product model in (10) yields a K distribution. In fact, if
the RCS pdf is a C distribution, that is
o o v o -1 ov
p (v) = ` v j C (o) e - vr ,
r
(11)
where o is an order parameter and vr is the mean, then the
pdf of the observed intensity signal is given by
2
Lo
p (I) = C (L) C (o) a k
rI
L +o
2
I
L +o -2
2
K o -L c 2
oLI m
rI ,
(12)
where K n ($) is the modified Bessel function of order n
and rI is the mean of intensity. Fitting the parameters of
the pdf to the observed signal allows information on the
RCS to be retrieved.
The model in (12) yields a pixelwise statistical description of the observed intensity values. A complete description
of the scene, however, needs the inclusion of the autocorrelation function into the model. If such a function is estimated from the observed data, then the exact autocorrelation function of the RCS is quite difficult to achieve and
usually it does not exist in a closed form [4].
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
III. SPACE AND SCALE-SPACE DOMAIN ESTIMATION
From the previous discussion, it emerges that modeling the received SAR signal should take into account
several physical, statistical and engineering aspects of
the overall system. Such a complexity makes the process of extracting average backscatter information from
the observed signal a nontrivial task. From a signal
processing perspective, a first step towards finding efficient solutions is stating the acquisition model in the
simplest form as possible. In [20], several multiplicative
models of speckle are described and classified according to the autocorrelations of the imaged scene and of
the noise term.
In the following of this section, models of the noisy signal in both spatial and transformed domains are reviewed,
Bayesian estimation principles are briefly recalled and the
wavelet transform, in both decimated and undecimated
versions, is introduced as a transformation suitable for
despeckling. Eventually, the modeling of pdfs for Bayesian
estimation in the wavelet domain is discussed and shown
to be crucial for performances.
A. MODELS OF NOISY SIGNAL
Perhaps, the most used model in the literature on despeckling is the following:
g = fu ,
(13)
where v = (u - 1) f accounts for speckle disturbance in an
equivalent additive model, in which v, depending on f, is a
signal-dependent noise process.
A second way that allows the multiplicative noise to be
transformed into an additive one is using a homomorphic
transformation [32]. It consists of taking the logarithm of
the observed data, so that we have
log g = log f + log u
g l = f l + ul ,
(15)
where g l , f l and ul denote the logarithm of the quantities in (13). Unlike the case in (14), here the noise
component ul is a signal-independent additive noise.
However, this operation may introduce a bias into the
denoised image, since an unbiased estimation in the logdomain is mapped onto a biased estimation in the spatial domain [33]; in math form, if u exhibits E [u] = 1,
E [ul ] = E [log (u)] ! log (E [u]) = log (1) = 0.
Over the last two decades, approaches to image denoising that perform estimation in a transformed domain have
been proposed. Transforms derived from multiresolution signal analysis [34], [35], such as the discrete wavelet
transform (DWT), are the most popular in this context.
Despeckling in a transform domain is carried out by taking
the direct transform of the observed signal, by estimating
the speckle-free coefficients and by reconstructing the filtered image through the inverse transform applied to the
despeckled coefficients.
where f is a possibly autocorrelated random process
and represents the noise–free reflectivity; u is a possibly autocorrelated stationary random process, indeB. BAYESIAN ESTIMATION CONCEPTS
pendent of f, and represents the speckle fading term; g
From the previous discussion about the most widely used
is the observed noisy image. All the quantities in (13)
signal models for despeckling, it can be seen that the multimay refer to either intensity or amplitude as well as to
plicative model is often manipulated in order to obtain an
single-look or multilook images, whose pdfs have been
additive one. Fig. 2 summarizes the various versions of the
described previously.
additive models.
The variable u may be assumed as spatially correlated [30].
The block “Estimator” attempts to achieve a speckleRecently, it has been shown [31] that a preprocessing step that
free representation of the signal in a specific domain; for
makes speckle uncorrelated, that is “whitens” the complex
signal, allows despeckling algorithms
designed for uncorrelated speckle to be
successfully applied also when speckle
is (auto)correlated. Therefore, in the
g=f+v
Estimator
fc
following we shall analyze only algo(a)
rithms working under the hypothesis of
exp
g = fu
Estimator
log
uncorrelated speckle.
fc
The nonlinear nature of the rela(b)
tionship between observed and
g=f+v
w -1
w
Estimator
fc
noise-free signals makes the filtering
procedure a nontrivial task. For this
(c)
reason, some manipulations have
exp
log
g = fu
Estimator
w
w -1
fc
been introduced to make the observa(d)
tion model simpler. Several authors
adopt the following model, derived
from (13):
FIGURE 2. Additive models commonly used in despeckling algorithms: (a) signal-dependent
in spatial domain, (b) signal-independent in spatial domain, (c) signal-dependent in transg = f + (u - 1) f = f + v,
(14) form domain, and (d) signal-independent in transform domain.
SEPTEMBER 2013
11
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
The estimate in (17) would require the knowledge of the
nonstationary joint pdfs of any orders.
A simpler solution requiring only second order moments
is the linear MMSE (LMMSE) estimator, in which the MMSE
solution is sought by constraining the estimator to be a linear combination of the observed variables. The LMMSE
estimator is given by
example, in the transform domain, as in Fig. 2-(c), or in the
homomorphic-transform domain, as in Fig. 2-(d), in which
the noise-free informative signal is contaminated with
additive signal-dependent or signal-independent noise,
respectively.
The basics of Bayesian estimation are now reviewed for
the simplest case, shown in Fig. 2-(a), even though analogous derivations hold for all the other cases in Fig. 2.
A Bayesian estimator [36] tries to achieve an estimate
tf of f —which is assumed to be a random process—by
having some prior information about the signal to be estimated, summarized in p F ( f), the a-priori pdf of f. Different
Bayesian estimators can be defined according to the choice
of the Bayesian “risk”, i.e., the function of the estimation
error f = f - tf we would like to minimize.
The minimum mean square error (MMSE) estimator
minimizes the quantity E [f 2] = E [( f - tf ) 2]. It is wellknown [36] that the solution is given by
tf MMSE = E F G [ f
g ],
tf LMMSE = E [ f ] + C fg C -gg1 (g - E [g]),
in which C fg is the covariance matrix between f and g and
C gg is the autocovariance matrix of g. Prior knowledge is
now embedded in the second order statistics of the noisefree and noisy signals, which can be derived by exploiting
the additive model.
The maximum a-posteriori probability (MAP) estimator minimizes the quantity E [C (f)], where C (f) = 1 for
| f | > d and C (f) = 0 elsewhere. The solution, when d is
small, is given by
(16)
tf MAP = arg max p F G [ f
which is the expectation of the noise-free signal conditional
to the noisy observation. By exploiting the Bayes rule and
the additive signal-dependent model g = f + v, we obtain
ft MMSE =
#
f
g ].
(19)
Again, by exploiting the Bayes rule and the additive model,
we have
fp F G ( f g) df
tf MAP = arg max p G F (g f ) p F ( f )
f
p G F (g f ) p F ( f )
df
f
p G (g)
#
# fp V (g - f ) p F ( f ) df
=
.
# p V (g - f ) p F ( f ) df
=
(18)
= arg max p V (g - f ) p F ( f )
(20)
f
(17)
or, equivalently,
tf MAP = arg max
f
Analysis Stage
x[n]
H1(z)
2
H0(z)
2
[log p V (g - f )
+log p F ( f )]
Synthesis Stage
H1(z)
2
2
G1(z)
H0(z)
2
2
G0(z)
+
2
G1(z)
2
G0(z)
xc[n]
+
(a)
Analysis Stage
x[n]
H1(z)
H0(z)
Synthesis Stage
G1(z)
2
2
H1(z2)
4
4
G1(z2)
H0(z2)
4
4
G0(z2)
+
+
xc[n]
G0(z)
(b)
FIGURE 3. (a) Two-level nonredundant wavelet decomposition/reconstruction and (b) the equivalent scheme obtained applying the noble identities. The undecimated wavelet transform is obtained
by eliminating the downsamplers and upsamplers contained in the shaded box.
12
(21)
Eqs. (17), (18) and (20) reveal
that all solutions, besides to
the a-priori information on f,
require also knowledge of the
pdf of the noise component v.
C. WAVELET TRANSFORMS
A Bayesian estimation carried
out in the spatial domain leads
to a solution that adaptively
depends on local statistics, i.e.,
is a space-adaptive estimator. A
Bayesian estimation carried out
in the multiresolution, or scalespace, domain may have the
extra value of leading to a scalespace adaptive estimator, that
is, an estimator adaptive not
only in space but also in scale.
Such an extra value is not automatic and requires careful pdf
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
j
H eq, l (z) =
H
j
eq, h
j -1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Normalized Frequency
FIGURE 4. Equivalent filters frequency responses obtained from
8-tap Daubechies orthogonal wavelets [34].
r
~y
HH
LH
HH
HL
LL
HL
-r
0
HH
LH
m =0
j -2
(z) = = % H 0 (z )G $ H 1 (z
~x
HH
(a)
Lowpass
Subband
x(n)
Bandpass
Directional
Subbands
Bandpass
Directional
Subbands
m
2m
r
-r
% H 0 (z 2 ),
(b)
2 j -1
),
(22)
m =0
where the subscripts l and h refer to the approximation (low
pass) and detail or wavelet (bandpass and high pass) signals, whereas j denotes the level of the decomposition. An
example of the equivalent filters frequency responses, relative to a four level decomposition, is shown in Fig. 4.
SEPTEMBER 2013
1
Magnitude
modeling in the transformed domain, otherwise the spatial
adaptivity may get lost in favor of the scale adaptivity.
The wavelet analysis provides a multiresolution representation of continuous and discrete-time signals and
images [35]. For discrete-time signals, the classical maximally decimated wavelet decomposition is implemented
by filtering the input signal with a low pass filter H 0 (z) and
a high pass filter H 1 (z) and downsampling each output by
a factor two. The synthesis of the signal is obtained with a
scheme symmetrical to that of the analysis stage, i.e., by
upsampling the coefficients of the decomposition and by
low pass and high pass filtering. Analysis and synthesis filters are designed in order to obtain the perfect reconstruction of the signal and by using different constraints (e.g.,
orthogonal or biorthogonal decomposition, linear phase
filters). Applying the same decomposition to the low pass
channel output yields a two-level wavelet transform: such
a scheme can be iterated in a dyadic fashion to generate a
multilevel decomposition. The analysis and synthesis stages
of a two-level decomposition are depicted in Fig. 3-(a).
In several image processing applications, e.g., compression, the DWT is particularly appealing since it compacts
energy in few coefficients. However, for most of the tasks
concerning images, the use of an undecimated discrete
wavelet transform (UDWT) is more appropriate thanks
to the shift-invariance property. UDWT is also referred
to as stationary WT (SWT) [37], [38], as opposite to Mallat’s octave (dyadic) wavelet decomposition DWT [35],
which is maximally, or critically, decimated. The rationale
for working in the UDWT domain is that in DWT, when
coefficients are changed, e.g., thresholded or shrunk,
the constructive aliasing terms between two adjacent subbands are no longer canceled during the synthesis stage,
thereby resulting in the onset of structured artifacts [39].
As to the construction of the UDWT, it can be shown
that if we omit the downsamplers from the analysis stage
and the upsamplers from the synthesis stage, then the
perfect reconstruction property can still be achieved. The
relative scheme for a two-level decomposition is depicted
in Fig. 3-(b). In the block diagram, by applying the noble
identities [40], the downsamplers (upsamplers) have been
shifted towards the output (input) of the analysis (synthesis) stage. Eliminating these elements yields the UDWT. As
a consequence, the coefficients in the transform domain
can be obtained by filtering the original signal by means of
the following equivalent transfer functions:
FIGURE 5. Frequency splitting from a single-level separable DWT (a),
obtained by low pass (L) and high pass (H) filtering along the rows
and the columns (LL, HL, LH, and HH denote all possible combination); in (b), the splitting obtained from the nonsubsampled Laplacian
pyramid decomposition (on the left) and the nonsubsampled directional filter banks (on the right) composing the contourlet transform.
13
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Let A xj (n) and W xj (n) denote the approximation and
wavelet coefficients, respectively, of the signal x at the jth
level of the decomposition, whereas n is the spatial index.
Since the wavelet transform is linear, from equation (14)
we have
j
A gj (n) = A f (n) + A vj (n)
j
W gj (n) = W f (n) + W vj (n).
(23)
(24)
Usually, only the wavelet coefficients (24) are processed
for despeckling; the baseband approximation is left
unchanged.
The wavelet transform is usually implemented for
images by using separable filtering along the columns and
the rows of the image. The effect of this processing is the
extraction, in each subband, of a rectangular region of the
frequency plane which corresponds, in the spatial domain,
to the extraction of horizontal and vertical details with different degrees of resolution. The frequency plane splitting
relative to a single level decomposition is given in Fig. 5-(a).
However, extracting directional information has been demonstrated to be useful in several image processing tasks.
Recently, multiresolution transforms embedding directional information, such as contourlets [41], curvelets [42],
[43], and many others, have been successfully applied to
denoising in general and despeckling in particular. The
nonsubsampled contourlet transform is a combination
of a nonsubsampled Laplacian pyramid (NLP) decomposition and of nonsubsampled directional filter banks
(NDFB). The relative frequency splitting is depicted in
Fig. 5-(b). As in the case of the UDWT, also the coefficients of the nonsubsampled contourlet transform can
be achieved by means of linear time-invariant (LTI) systems directly applied to the input, which allows statistical
parameters to be easily computed. Using directional information is effective in terms of despeckling performance
[44], even though a higher computational cost must be
paid due to the need of a nonseparable implementation.
By assuming that the transform is linear, the additive models in (14) and (15) can be easily generalized to
the transformed domain. Specifically, for the formulation
given in (14), if W x denotes the transform operator applied
to the signal x, we have
W g = W f + Wv .
(25)
In an analogous way, by applying both the homomorphic
filtering concept and the linear transform, the observation
model in (15) becomes
W gl = W f l + W ul .
(26)
The Bayesian estimator explicitly derived for the additive model in (14), can also be applied to the additive
models defined in (15), (25), and (26) by simply changing the type of variables and prior knowledge, that is: 1)
the prior pdf of the signal of interest (related to the reflec14
tivity) and represented by f, f l, W f and W f l in equations
(14), (15), (25), and (26), in that order; 2) the pdf of the
additive noise component, represented by v, ul , W v and
W ul, in the same equations.
D. PDF MODELING
Bayesian estimation relies on an accurate probabilistic
modeling of the signals under concern. However, the choice
of pdfs suitable for modeling the data of interest is not a
simple task. In Section II, we have described some of the
most used pdfs for the speckle and reflectivity processes.
While the former derive from the image formation mechanism and may be considered as valid in most of the images
where the fully developed speckle model holds, the latter
highly depend on the imaged scene. We highlight again
that different types of landscapes and land covers induce
different distributions on the reflectivity signal. Models of
the underlying land cover may help to derive a pdf of the
imaged signal, but this knowledge may not be available for
despeckling or may be insufficient. As to the modeling of
signals in the homomorphic domain, an exact derivation
of the log-intensity and of the log-amplitude of the fading
variable is available [33], whereas the characterization of
the backscattering coefficient still remains crucial.
The modeling of the involved variables may be simpler
and more robust if one works in a multiresolution, or scalespace, domain, instead than in the spatial domain. In fact,
it is well-known that the pdf of wavelet coefficients can
be approximated by several unimodal distributions—as
noticed by Mallat in his seminal paper [35], where a generalized Gaussian was used—that can be described by a small
number of parameters. They can be adaptively estimated
from the coefficients of the observed image, independently
of the distribution of the image that is transformed.
Validating a hypothetical pdf model is, in general, quite
hard. In some works, wavelet coefficients pdfs are validated
“globally” from the observation of the histogram of the
amplitude of the coefficients in a whole subband. However,
since the signal is nonstationary, spatially adaptive methods should be used instead. A single observation, or realization, of the scene is usually available; thus, one may only
conjecture that wavelet coefficients “locally” follow a given
distribution (only few samples are available to perform the
validation of the local model) whose parameters locally
vary. A way to check the validity of the pdf model is experimentally observing the performances of despeckling filters
on either true SAR or synthetically speckled images. As a
general rule of thumb, the higher the number of parameters, or degrees of freedom, of the pdf, the better its ability
to model the true wavelet coefficients pdf within a whole
subband, but the lower their estimation accuracy from the
few samples available in a local window within a subband
and the higher the complexity of the resulting estimator.
Therefore, the use of reasonably simple distributions may
be expected to yield better results than more complex ones,
that is, overfitting is not rewarding.
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Another fact that should be considered when a pdf
model is chosen is the computational cost. Some combinations of estimation criterion and pdf model yield a Bayesian
estimator that can be achieved only numerically [45]. This
fact may prevent from using the filter when huge amounts
of data need to be processed. In this case, a closed form
solution may be preferred, even though a possible loss of
performances may be experienced.
IV. A REVIEW OF DESPECKLING METHODS
A multitude of despeckling filters can be obtained by
combining the different domains of estimation (spa-
tial, homomorphic, wavelet, homomorphic-wavelet),
the estimation criteria, e.g., MMSE, LMMSE, minimum
mean absolute error (MMAE), MAP or non-Bayesian,
and the pdf models. A nonexhaustive review and classification of methods is attempted in the following of
this section.
A. BAYESIAN METHODS IN SPATIAL DOMAIN
Early works on despeckling were deployed in the spatial
domain and were obtained by making assumptions on the
statistical properties of reflectivity and speckle, e.g., pdf
and autocorrelation function.
(a)
(b)
(c)
(d)
FIGURE 6. Examples of the application of Bayesian estimators in the UDWT domain: (a) original 5-look ERS-2 image and filtered versions
obtained with, (b) Lee refined filter [49], (c) refined C-MAP filter [56], and (d) Rational Laplacian Pyramid filter [59].
SEPTEMBER 2013
15
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Lee Filter—The local-statistics filter, introduced by
Jong-Sen Lee in 1980, is reportedly the first model-based
despeckling filter. The original paper [21] contained solutions for both additive signal-independent noise and
speckle noise. The latter solution was thoroughly developed in [46] and reviewed in [47] together with the sigma
filter. An LMMSE solution was derived by linearizing the
multiplicative noise model around the mean of the noisy
signal. In this way, the author devised an approximate but
effective solution which is identical to the exact solution,
apart from the term (1 + v 2u) -1, in which v 2u is the variance
of the multiplicative noise u. The contribution of this term
can be practically neglected for multi-look images, in which
v 2u % 1 [19], [48].
Lee Refined Filter—This filter [49] was designed to overcome the drawback of edge boundaries that are left noisy by
Lee filter. To improve filtering, once an edge is detected in a
7 # 7 sliding window, the algorithm uses the local gradient
to estimate its orientation. Eight edge-directed non-square
windows are allowed. The estimation of the local mean and
of the local variance are performed within the local window
that better fits the edge orientation. If no edge is detected,
the estimates are computed on the whole 7 # 7 window.
Filtering results are quite impressive, particularly on edges
and high contrast areas. Some artifacts may occur when the
filter processes textured areas that result to be overly segmented. Another limitation is that the filter works with a
window of fixed size 7 # 7: textures characterized by a high
spatial variation and thin linear features may be altered. An
ERS-2 image of Florence is shown in Fig. 6-(a); processing of
refined Lee filtering in Fig. 6-(b).
Frost Filter—In Frost filter [50], starting from a model
of the coherent imaging system, a parametric approximation of the autocorrelation function of reflectivity is derived
from local statistics. Such a function is used to devise an
LMMSE solution for the noise-free reflectivity itself. The filtered value is a linear combination of pixel values within
the local window with a Gaussian weighting function that
depends on the local coefficient of variation of the noisy
image g, namely C g, defined as the ratio of local standard
deviation to local mean. Despite its large popularity in the
image processing community, Frost filter had no developments over time, either by the authors or by others, apart
from the heterogeneity adjustment common to all spatial
Bayesian filters [51], which will be reviewed at the end of
this subsection.
Kuan Filters—Kuan’s filter [52] exactly implements the
LMMSE solution (18) starting from a signal model that features nonstationary mean, nonstationary variance and thus
a diagonal covariance matrix in (18). The resulting LMMSE
solution is referred to as local LMMSE (LLMMSE) to indicate that it contains only local first order statistics, mean
and variance, that are easily calculated in a sliding window. Accordingly, the optimum estimate of reflectivity, tf,
is given as a combination of the unfiltered noisy pixel value
g and of its local average gr, approximating the local mean,
16
with weights nonnegative and summing to one. The center
pixel is more or less weighted depending on its local signal
to noise ratio (SNR). Besides despeckling, also restoration
for the effects of the imaging system can be carried out [53].
The price is a considerable increment in the computational
complexity of the procedure.
MAP Filters—The prototype of MAP filters in spatial
domain is the C-MAP filter, introduced in [54] and thoroughly analyzed in [55]. It assumes that both the radar
reflectivity and the speckle noise follow a Gamma distribution and solves the MAP equation (21) accordingly. It is
designed to smooth out noise while retaining edges or shape
features in the image. Different filter sizes greatly affect the
quality of processed images. If the filter is too small, the
noise filtering algorithm is not effective. If the filter is too
large, subtle details of the image will be lost in the filtering
process. A 7 # 7 filter usually gives the best tradeoff.
A refined version of the C-MAP filter that features
an improved geometrical adaptivity, analogously to
Lee refined filter, was proposed in [56]. The visual result
appears in Fig. 6-(c). This achievement marks the beginning
of a certain performance saturation in spatial despeckling
methods, although highly sophisticated Bayesian methods
in space domain, featuring MAP estimation associated to,
e.g., Gauss-Markov and Gibbs random fields for prior modeling have been introduced later [57] and are still used [58].
Despeckling Filters and SAR Heterogeneity—The
filters described in this subsection can be adjusted to the
heterogeneity of SAR images, as suggested in [51]. The
rationale is that in true SAR images at least three statistical classes can be recognized: homogeneous, textured, and
strong, or persistent, scatterer. The first class is characterized by a spatially constant reflectivity and in this case the
best estimator is a plain average of intensity pixel values in
a neighborhood. Pixel belonging to the third class should
be detected and left unprocessed, as they are intrinsically noise-free and are used for calibration, registration,
etc. The intermediate class may be processed through the
desired filter, e.g., Lee, Frost and Kuan filters. The resulting filters are known in the literature as enhanced Lee,
Frost and Kuan filters [51]. The C-MAP filters was originally defined in enhanced version [54]. The three classes
are found by thresholding C g . The two thresholds, namely
C min and C max are empirically set equal to v u, the standard
deviation of speckle, and 3 v u [51].
B. BAYESIAN METHODS IN TRANSFORM DOMAIN
Apart from a few methods that employ multiresolution
concepts without a formal multiresolution analysis, like
Meer’s filter and especially the filter based on the Laplacian
pyramid, all filters reviewed in this subsection exploit the
discrete wavelet transform, either decimated or not.
Meer’s Filter—Meer’s filter [60] considers a local neighborhood constituted by a set of three concentric sliding
windows, 7 # 7, 5 # 5, and 3 # 3. A homogeneity index
is given by C g, computed over each of the windows. The
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
the reflectivity signal and to a Gaussian pdf for the noise;
spatial average on the largest window satisfying a homogethe previous method has been extended to the MAP criteneity criterion, defined by thresholding its C g, is given as
rion in [75]. In [76], MAP estimation is used associated to
output. If such a window does not exist, Kuan’s LLMMSE
a heavy-tailed Rayleigh prior for the signal and to Gamma
estimate on the innermost 3 # 3 window is assigned to the
or Nagakami models (for images in intensity or amplitude
center pixel. This filter is effective in preserving point tarformat, respectively) for the noise.
gets, linear features and edges, thanks to its
Non-Homomorphic Filtering in Wavecapability to shrink its window size. Perlet
Domain—Nonhomomorphic waveletformances on point targets and linear feaDESPECKLING PURSUES
domain
despeckling (see Fig. 2-(c)) has been
tures are slightly better than those of Lee’s
THE TECHNOLOGICAL
considered
less frequently in the literature.
refined filter which, however, is superior on
OBJECTIVE OF GIVING AN
Even though the absence of the bias due to
linear edges.
EXTRA VALUE TO THE
the nonlinear mapping of the logarithm is an
RLP Filter—The rational Laplacian pyrHUGE AMOUNT OF DATA
advantage, the estimation of the parameters
amid (RLP) filter [59] is an evolution for
ROUTINELY COLLECTED
of the signal and noise pdfs becomes more
speckle filtering of the denoising method
BY SATELLITE SAR
complex. In fact, in the equivalent additivebased on the enhanced Laplacian pyramid
SYSTEMS.
noise model for the non-homomorphic case,
[61]. The latter is commonly used for spathe noise term is signal-dependent and, theretially scalable layered video coding as well
fore, its parameters are much more difficult to
as for lossless and near lossless compresbe estimated.
sion of still images by exploiting quantization noise feedIn the seminal paper by Foucher et al. [77], undeciback [62], [63].
mated wavelet was firstly used for despeckling. EstimaRLP differs from LP because its passband layers are
tion is based on the MAP criterion and the Pearson system
obtained by taking the ratio pixel by pixel between one
of distributions. In [78], the LMMSE estimator, optimal
level of the Gaussian pyramid and the interpolated version
under the Gaussianity assumption, has been presented.
of the lower resolution upper level. While the baseband
In [79], the LMMSE estimator with mixtures of Gaussian
icon, corresponding to the top of the Gaussian pyramid,
pdfs is enforced by the use of the ratio edge detector [80] to
may be left unprocessed because of its high SNR obtained
improve despeckling of contours. In [81], MAP estimation
through cascaded low pass filtering and decimation stages,
is used along with the assumption of normal inverse Gaussanalogously to multi-looking, the bandpass levels of RLP
ian distributions for the wavelet coefficients. MAP estimaare processed by means of Kuan’s filter [52]. The despecktion associated to locally varying generalized Gaussian
led image is synthesized from the denoised RLP. This mul(GG) distributions has been used in [82]. In [83] a segmentiscale LLMMSE filter outperforms its spatial counterpart
tation-based MAP despeckling with GG priors is achieved.
thanks to multiresolution processing. The result of RLP filThe method in [82] has been extended to the domain of
tering can be watched in Fig. 6-(d).
the nonsubsampled contourlet transform (NSCT) in [44].
Homomorphic Filtering in Wavelet Domain—FilterAnother method in the contourlet domain has been proing in the wavelet-homomorphic domain (see Fig. 2-(d))
posed in [84]. MAP and MMSE estimators associated to
has been extensively used during the last twenty years and
Laplacian and Gaussian PDFs for the signal and noise compotentially superior performances over conventional spaponents have been proposed in [85]. Generalized C and
tial filters have been recognized [64], [65]. In fact, each
wavelet subband is associated to a speckle contribution that
Gaussian distributions have been used for MAP despeckmay be exactly measured [66] and filtered out. Thus, spaling in [86], [87]. An interesting example of despeckling
tially adaptive filtering become also scale-adaptive.
not pertaining SAR but ultrasound images and based on
Classical hard- and soft-thresholding methods [67]
statistical classification of signal/noise wavelet coefficients
were applied in [68]. Thresholding based on nonlinear
is presented in [88]. Also non-Bayesian methods based on
functions (sigmoid functions), adapted for each subband,
the classification of signal and noise wavelet coefficients
has been used in [69]. In [70], MMSE estimation has been
have been proposed in [89], [90].
used associated to a combination of generalized Gaussian
(GG)/Gaussian pdfs for the reflectivity and for the noise
C. NON-BAYESIAN APPROACHES
components, respectively. In [71], MMSE estimation have
A number of despeckling filters published over the last
been used after modeling wavelet coefficients by means of
thirty years do not follow a Bayesian approach. In the folGaussian mixtures and Markov random fields to characlowing, the most popular approaches and related methods
terize their spatial and interscale dependency. In [72], the
are summarized.
MAP criterion has been used associated to a-stable disOrder Statistics and Morphological Filters—Starting
with median filter, order-statistics filters encountered a certributions for the prior of the signal and to a log-normal
tain popularity for despeckling, thanks to their peculiar feapdf for the noise. In [73], MAP estimation is applied based
ture of edge preservation. A conditional version of median
on normal inverse Gaussian distributions. In [74], MMAE
filter [91], replaces the central pixel value of the local sliding
estimation has been used associated to a Cauchy prior for
SEPTEMBER 2013
17
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
window with the sample median if and only if the former
is recognized as an outlier, i.e., an extremal value within
the window. An adaptive version of the weighted median
filter was specifically proposed for despeckling [92]. It is
substantially a center weighted median filter, in which the
weight is adaptively calculated from local statistics, in order
to preserve edges, retain textures and smooth the noisy
background.
Geometric filter (GF) [93] is a powerful tool for edgepreserving smoothing of noise and especially of speckle,
the purpose it was designed for. GF iteratively erases noise
samples regarded as geometric artifacts of the 3-D shape
defined by the 2-D gray-level function. GF is a nonlinear
local operator that exploits a morphologic approach to
smooth noise one image line at a time using a complementary hull algorithm, whose iterations converge towards
roots, i.e., steady patterns invariant to further iterations,
in which spatial details thinner than a critical size are
completely suppressed. Thicker objects are just slightly
smoothed and therefore fairly preserved as filtering is iterated. A decimated version of GF [94], suitable for spatially
correlated noise, including speckle, consists of applying GF
to the four polyphase components, in which the original
image is preliminarily decomposed, and of reinterleaving
the filtered components to yield the denoised image.
None of the methods reported above explicitly accounts
for the speckle noise model. However, their computational
speed together with the capability of preserving abrupt discontinuities of level was found to be of interest in computer
vision and object recognition applications. Nowadays these
methods are less and less frequently used, though they
might be valuable for high-resolution images of man-made
environments, in which persistent scatterers and not fully
developed speckle are frequently encountered.
Anisotropic Diffusion—Anisotropic diffusion [95] is
a technique, extremely popular in the image processing
community, that aims at reducing image noise without
removing significant parts of the image content, typically edges, lines or other details that are important for
the interpretation of the image. The derivation speckle
reducing anisotropic diffusion (SRAD) is tailored to
coherent images [96]. SRAD is the edge-sensitive diffusion for speckled images, in the same way that conventional anisotropic diffusion is the edge-sensitive diffusion for images corrupted with additive noise. Just as the
Lee and Frost filters utilize the coefficient of variation in
adaptive filtering, SRAD exploits the instantaneous coefficient of variation, which is shown to be a function of the
local gradient magnitude, and Laplacian operators. SRAD
overcomes traditional speckle removal filters in terms of
mean preservation, variance reduction, and edge localization. However, the unrealistic smoothness introduced
after iterated processing makes SRAD unsurpassed for
cartoon-like images, i.e., made up by textureless geometric patches, but may be unsuitable for real SAR images,
because fine details and textures that may be useful for
18
analysis are destroyed. A notable application of SRAD is
for coastline detection in SAR images [97].
Simulated Annealing Despeckling—Simulated Annealing (SA) was originally used for SAR image despeckling and
segmentation by White [98]. SA is a stochastic optimization
method used for finding the global maximum of an a-posteriori multivariate distribution, or equivalently the global
minimum of a multidimensional energy function, which
is often made difficult by local maxima (minima), which
can easily trap the optimization algorithm. SA is iterative
by nature, where a new configuration for iteration is found
from the previous configuration by applying a generation
mechanism and accepting the new configuration using
an acceptance criterion based on the energy divergence.
The temperature variable controls the optimization, and it
is decreased throughout the optimization process. For the
first iterations, when it is high, there is a high probability
of accepting configurations resulting in an increase in the
energy, thus making SA able to get out of local minima. As
the temperature is gradually decreased, the probability of
accepting configurations resulting in increasing energies is
reduced, so that at the end of the minimization no increases
are accepted, and the global minimum configuration is ideally reached. Despite its potentiality, the unlikely cartoonlike smoothness produced by SA was noticed in [99]. After
that, SA was used only in conjunction with complex multivariate pdf models, like in polarimetric SAR [100].
Sigma Filter—A conceptually simple noise smoothing
algorithm is the sigma filter originally developed for additive signal-independent noise [101] and promptly extended
to speckle removal [102] also in a comparison with local
statistics filtering [47]. This filter is motivated by the sigma
probability of the Gaussian distribution, and it smooths the
image noise by averaging only those neighborhood pixels
which have the intensities within a fixed sigma range of the
center pixel. Consequently, image edges are preserved, and
subtle details and thin lines, such as roads, are retained.
An enhanced version of Lee’s sigma filter [103] is derived
and proposed for unbiased filtering of images affected by
multiplicative noise with speckle statistics. Instead of the
plain point value, a more accurate start value is first produced, and then fed to the procedure of conditional average. A robust estimate of the nonstationary mean is defined
according to a decision rule. The start value is provided by
a nonlinear decision rule, aimed at rejecting noisy samples, that is performed on the averages computed within
four isotropically balanced pixel sets able to capture step
edges and thin lines. The level range of pixels to be averaged, adaptively defined as the product of the space-variant
mean estimate by the constant noise variance, is also forced
to account for the imbalance of the noise distribution, for
unbiased processing.
Eventually, in [104] the bias problem is solved by redefining the sigma range based on the speckle pdf. To mitigate the problems of blurring and depressing strong reflective scatterers, a target signature preservation technique is
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
bilistic measure that takes into account the pdf of SAR
developed. In addition, the LLMMSE estimator for adaptive
data, and by proposing an iterative procedure for refining
speckle reduction [21], [52] is incorporated.
the weights. Following a similar approach, an improved
Bilateral Filtering—The bilateral filter (BF), originally
similarity measure has been recently proposed in [112].
introduced in [105] for gray scale images, has been recently
Other approaches consider a Bayesian NL framework
extended to despeckling in [106]. The rationale of BF is that
[113], which has been applied to the despeckling of both
each pixel value within a sliding window is weighted both
ultrasound images [114] and SAR images [115]. The NL
for the distance to the center, as in Frost filter, and for the
principle has been successfully applied also to despeckdifference to the value of the center, as in sigma filter. In
ling in the wavelet domain [109], [116]. Namely, in [109]
an adaptive version of BF suitable for despeckling [107],
the authors extend the BM3D filter by redefining the
the spatial weighting is a Gaussian function, whose span
similarity measure among block of pixels according to
depends on the local coefficient of variation, analogously
[108], and employing the LMMSE principle [78] in the
to the enhanced Frost filter. A rule borrowed from [108]
estimation step.
defines the weights as the gray level difference between the
Total Variation Regularization—Another popular
central pixel and each neighboring pixel, as the probabildenoising approach is based on total variation (TV) reguity of two values in a speckled image that exhibit the same
larization [117]. In such a method, denoising is achieved
reflectivity value. The adaptive method in [107] exploits an
through the minimization of a suitable cost function,
order statistic filter, like [91], to reject outliers that often
combining a data fidelity term with a prior that enforces
occur. Despite its elegance and relatively low computasmoothness while preserving edges. Several solutions
tional cost, in the presence of strong noise, like for singleexist to apply TV methods in the case of multiplicative
look images, speckle-oriented BF suffers from limitations
noise [118]–[124]. These solutions differ according to the
given by the finite size spatial function, same as all local
domain in which the optimization is performed, which
spatial filters. A way to overcome such a drawback is adoptcan be either the intensity or the logarithm of the intening a nonlocal filtering approach.
sity, and the definition of the data fidelity term. In [119],
Nonlocal Filtering—Among the despeckling methods
the authors define the optimization problem in the
that cannot be included in the classical Bayesian frameoriginal intensity domain and apply a data fidelity term
work, nonlocal (NL) filtering is surely one of the most interbased on a maximum a posteriori approach, assuming
esting and promising solutions [108], [109]. NL filtering is a
a Gamma distributed speckle and a Gibbs prior. Due to
generalization of the concept of data-driven weighted averthe difficulty of defining strictly convex TV problems in
aging, in which each pixel is weighted according to its simithe original intensity domain, several authors have conlarity with the reference pixel, as in the pioneering sigma
sidered the logarithmic domain instead. When applying
filter. The NL mean filter [110] extends the above method,
TV regularization in the logarithmic domain, convex
by defining the weights as a function of the Euclidean disTV problems can be obtained by applying different data
tance between a local patch centered at the reference pixel
fidelity terms, including the L 2 norm [120], MAP based
and a similar patch centered at a given neighboring pixel.
The block-matching 3-D
filter (BM3D) [111] combines the advantages of
Subband
Bayesian
the NL principle and of
UDWT Domain
Spatial
PDF
Estimation
Statistics
Statistics
the wavelet representaModeling
tion: 3-D groups of pixels
are formed by collecting
blocks of pixels drawn
from different image
Speckled
Subband
Inverse
Despeckled
UDWT
locations and chosen
Image
Processing
UDWT
Image
according to their similarity with a reference
block, and Wiener filterLegend:
ing is applied to the wavelet coefficients of such
“Necessary to” Relation
3-D groups.
Processing Flow
In [108], NL filterSubband
UDWT: Undecimated Wavelet
Processing
ing has been applied
Transform
to despeckling by subUDWT Domain
stituting the Euclidean
distance used in the NL
mean filter with a proba- FIGURE 7. Flowchart of Bayesian filtering in the undecimated wavelet domain.
SEPTEMBER 2013
19
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
on Gamma distributed speckle [120], [124], a combination of the previous terms [121], the generalized Kullback-Leibler divergence [123], the L 1 norm on curvelet
coefficients [122]. It is worth noting that all the above
methods have been mainly validated on simulated data.
The literature regarding the application of such methods to actual SAR images is quite scarce [125]–[127], and
there is a general lack of comparisons with Bayesian and
NL despeckling methods.
Despeckling Based on Compressed Sensing—A new
signal representation model has recently become very
popular and has attracted the attention of researchers
working in the field of restoration of images affected by
additive noise as well as in several other areas. In fact,
natural images satisfy a sparse model, that is, they can
be seen as the linear combination of few elements of a
dictionary or atoms. Sparse models are at the basis of compressed sensing [128], which is the representation of signals with a number of samples at a sub-Nyquist rate. In
mathematical terms, the observed image is modeled as
y = Ax + w, where A is the dictionary, x is a sparse vector,
such that ;x; 0 # K, with K % M, with M the dimension
of x, and w is a noise term that does not satisfy a sparse
model. In this context, denoising translates into finding the sparsest vectors with the constraint ;y - Ax; 22 < e,
where e accounts for the noise variance. The problem
is NP-hard, but it can be relaxed into a convex optimization one by substituting the pseudo-norm ; $ ; 0 with
; $ ; 1 . Recently, some despeckling methods based on the
compressed sensing paradigm and sparse representations
have appeared [129]–[131].
V. MULTIRESOLUTION BAYESIAN FILTERING
In this section, we review some methods recently proposed for despeckling in the undecimated wavelet domain
that use a multiresolution analysis. The methods refer to
the additive model in (26), that is, they do not exploit the
homomorphic transform, which may introduce bias in the
estimation of the despeckled image.
Fig. 7 outlines the flowchart of Bayesian despeckling
in UDWT domain. As it appears, the majority of processing is carried out in the transform domain. Statistics in the
transform domain are directly calculated from the spatial
statistics of the image by exploiting the equivalent filters
(4), as firstly proposed by Foucher et al. [77].
A. LMMSE FILTER
In the case of zero-mean Gaussian pdf modeling for the
quantities W f and W v, the MMSE and MAP Bayesian estimators are identical. The expression of the filter has a simple
and closed analytical form that depends only on the space
varying variance of the wavelet coefficients [78], that is
t LMMSE
= Wg $
W
f
2
vW
f
-1 -1
.
2 = W g $ (1 + SNR )
v + vW
v
2
Wf
Thus, LMMSE estimation corresponds to a shrinkage of the
noisy coefficient by a factor inversely related to its SNR.
Unfortunately, the wavelet coefficients of noise–free reflectivity do not respect the Gaussian assumption, especially
in the lowest levels of the wavelet decomposition, so that
its performance are inferior to more complex Bayesian estimators. In Fig. 10-(a) a single-look COSMO-SkyMed image
is shown. In Fig. 10-(b) the despeckled image obtained by
applying the LMMSE estimator is presented.
B. MAP FILTERS
In this section, we present two different filters that use the
MAP estimation criterion but different models for the pdfs
of the wavelet coefficients relative to the original reflectivity and to the additive signal–dependent noise.
Equation (21) can be rewritten as
t MAP
W
= arg max [ln p WV
f
Wf
3
2.5
o=0.5
o=1
o=2
o=3
pX (x)
2
1.5
1
0.5
0
-3
-2
-1
0
x
1
2
3
FIGURE 8. Zero-mean GG pdfs obtained with unity variance and
different os.
20
(27)
WF
(W g - W f W f ) + ln p WF (W f )].
(28)
Since the signal and noise processes are nonstationary,
space varying pdfs must be considered. The pdfs that are
considered here can be seen as a trade-off between simplicity (few parameters to be estimated from the observed data)
and modeling capability.
MAP–GG filter—In [82], the MAP criterion is combined with a generalized Gaussian (GG) distribution
for the wavelet coefficients. Since the birth of the wavelet recursive algorithm by Mallat [35], a GG pdf has been
used to model image wavelet coefficients and several other
authors use the GG distribution for many image processing tasks involving wavelets. A zero-mean GG pdf depends
only on two parameters and is characterized by being symmetric around the mean. Its expression is given by
o $ h (o, v) - [h (o, v) $| x |] o
Fe
,
2 $ C (1/o)
(29)
SEPTEMBER 2013
p X (x) = <
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
6
WfMAP
4
2
0
-2
-4
-6
-6
-4
-2
0
Wg
2
4
6
2
4
6
(a)
6
4
WfMAP
2
0
-2
-4
-6
-6
-4
-2
0
Wg
(b)
tf
FIGURE 9. Mapping of the W
MAP
estimates vs the observed W g: in
(a) v W f = 2, v Wv = 1, o Wv = 2 and o W f varies from 0.4 to 2 with
step 0.2; in (b) o Wv = 1.2 (the other parameters are unchanged).
where C is the Gamma function, v is the standard deviation of the distribution, v is a shape factor, and h (o, v) is
given by
1/2
1 C (3/o) F
.
h (o, v) = v <
C (1/o)
(30)
The GG distribution is reasonably simple, since the use of
only two parameters allows different levels of “peakedness”
to be achieved. As particular cases, the GG pdf includes
both the Laplacian and the Gaussian pdfs, for o = 1 and
o = 2, respectively. A plot of GG pdf curves for different
values of o is shown in Fig. 8.
Substituting (29) into (28) yields
hW f oW f
t MAP
W
= arg max ;ln 2C (1/o ) - ^h W f W f hoW f
f
Wf
Wf
h Wv o Wv
ln 2 (1/ ) - ^h Wv W g - W f ho WvE .
C o Wv
(31)
In [82], a method for the estimation of the parameters
relative to the GG model, i.e., the standard deviation v
and the shape factor v of the distributions relative to
W f and W v, is given. The estimation of the parameters
is based on the computation of some moments of the
observable variables g and W g . In the implementation
SEPTEMBER 2013
of the filters, these moments are substituted by spatial
averages. The solution of equation (31) is not known in a
closed analytical form and a numerical optimized solution has been proposed in [82].
t MAP
In Fig. 9, a set of curves plotting W
vs W g is given
f
for particular values of the parameters of the GG model: in
Fig. 9-(a), the curves refer to v W f = 2, v Wv = 1, o Wv = 2 and
to o W f varying from 0.4 to 2 with step 0.2; in Fig. 9-(b), the
parameter o Wv has been changed to 1.2 (the other parameters were not modified). Such curves define a remapping
of the observed coefficients onto noise-free ones same as it
is done by hard and soft-thresholding schemes commonly
used for denoising signals affected by additive signalindependent noise [67], [132]. It is important, however,
to point out that for despeckling the wavelet coefficients
are modified according to the multiplicative model of
speckle and thus adaptively vary according to the locally
estimated parameters.
MAP–LG filter—In [85], the empirical distribution
of the shape factor of noise–free reflectivity coefficients
has been investigated and an interesting behavior was
noticed. For the lowest levels of decomposition, the shape
factor is usually very close to one, whereas it tends to shift
towards two in highest ones. The shape factor of signal–
dependent noise coefficients, instead, are mostly concentrated around the value two. These facts suggest directly
introducing a combination of Laplacian and Gaussian
pdfs into the modeling: this yields some computational
advantages with respect to using the more general GG
pdf. In fact, by assuming that the wavelet coefficients W v
and W f follow a zero-mean Gaussian and zero–mean
Laplacian distribution, respectively, yields the following
closed form estimator [133]:
t MAP
W
= arg max p W f W g (W f W g)
f
Wf
Z W g - t,
if W g > t
]
if W g < - t
= [ W g + t,
]
otherwise
0
\
(32)
where t = 2 v 2Wv /v W f . Thus, the estimator is equivalent to
a soft-thresholding algorithm with a locally adaptive threshold. Eq. (32) has been originally devised in [134] and used
for processing ultrasound images with decimated wavelets.
C. ADJUSTMENTS FOR SAR IMAGE HETEROGENEITY
In several despeckling methods, different filtering strategies are used according to the texture content of the scene.
In [51], [54], the coefficient of variation is used to discriminate among homogeneous, textured and highly heterogeneous (or point target) areas. Pixel belonging to the
first two classes are filtered by using simple averaging and
C-MAP, or another local-statistics filter, whereas no filtering is attempted on point targets. A strongly scattering
target, however, is concentrated in space, but after wavelet analysis its response will be somewhat spread because
of the finite support of the wavelet function. Thus, also
21
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
(a)
(b)
(c)
(d)
FIGURE 10. Examples of the application of Bayesian estimators in the UDWT domain: (a) original COSMO-SkyMed 4-look StripMap image,
filtered versions obtained with (b) LMMSE, (c) MAP-GG with segmentation (GG–MAP–S), and (d) MAP-LG with classification (LG–MAP–C).
UDWT coefficients around a point target one pixel wide
will depend on the target response, unlike what happens
in space. In the past, this was perhaps the main objection
towards a systematic use of the wavelet transform to analyze SAR images. Starting from [83] a preprocessing step
of point targets, and thicker strong scatterers in general,
was devised. Targets are detected as upper percentiles of
the image histogram, removed from the image and stored.
Void pixels are smoothly filled by interpolating their
neighbors. Then, wavelet analysis is performed. After synthesis of the despeckled image, point targets are reinserted
in their original places.
22
The leftover two classes, namely homogeneous and
textured, can be handled also by multiresolution methods
to improve their performance. In [83], UDWT subbands
are segmented into texture classes according to an energy
index computed in the UDWT domain. Several classes of
texture, from textureless onward, can be recognized. The
wavelet coefficient of each segment on each subband are
supposed to have a unique shape factor of the GG function, while the variance is calculated for each coefficient.
Thus, the calculation of the v is more accurate than in [82],
thanks to the more consistent sample size. In [85], the segmentation has been extended to the MAP-LG filter. This
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
time there are no parameters to estimate on segments, as
for GG. Thus, a classified approach consists of switching
among different estimators, e.g., MAP-LG and LMMSE,
depending on the degree of texture of each segment. In
Fig. 10-(c) and 10-(d), the results of MAP-GG and MAP-LG
estimators, the former with segmentation (GG–MAP–S),
the latter with classification (LG–MAP–C), on the image
in Fig. 10-(a) are shown.
A segmentation based approach seems also a natural
solution to changes in the speckle model occurring as
the spatial resolution of single-look products increases.
This happens for very high resolution (VHR) new generation SAR systems, especially with Spotlight products. As
the size of the elementary resolution cell decreases, the
assumption of distributed scatterers is less and less verified. In substance, what is homogeneous at 10 m scale
may no longer be so at 1 m. So, we expect that VHR SAR
images are more textured and contain more persistent
scatterers, and less homogeneous regions, than earlier
products. A viable solution with segmented processing in
UDWT domain is introducing corrective factors for under
smoothing in textured segments, depending on the class
of texture energy measured in the UDWT domain, analogously to what proposed in [78].
VI. NON-LOCAL MEAN FILTERING
The NL mean (NLM) filter proposed by Buades et al. in
[110] is based on the simple idea of estimating the noise
free image as a weighted average of noisy pixels
tf (n) = / m w (n, m) g (m) ,
/ m w (n, m)
(33)
(a)
where the weights w ^n, m h take into account the “similarity” between pixels g ^n h and g ^m h . The key idea of
the NLM filter is that the weights w ^n, m h are based on
the Euclidean distance between local patches centered at
g ^n h and g ^m h, according to
1
w (n, m) = exp c - h / a k g (n + k) - g (m + k) 2 m ,
(34)
k
where a k ’s define a Gaussian window and h controls the
decay of the exponential function.
The NLM filter obtains a very good performance in the
presence of additive white Gaussian noise, since the Euclidean distance is a natural similarity measure for this kind of
model. However, in the case of SAR images, the weights have
to be generalized to the case of multiplicative and non-Gaussian speckle. It is also interesting to combine the effectiveness
of the NL principle with the benefits of the sparse representation offered by the wavelet transform. In the following, we
will review two SAR despeckling filters based on the NL principle in the spatial [108] and in the wavelet domain [109].
A. PROBABILISTIC PATCH-BASED FILTER
The probabilistic patch-based (PPB) filter, proposed by
Deledalle et al. in [108], extends the NLM filter to the
domain of SAR images by exploiting its connections to
the weighted maximum likelihood extimator (WMLE).
Namely, under the WMLE principle the noise-free image
can be estimated as the value maximizing a weighted likelihood function of the observed data
tf (n) = arg max / w (n, m) log p (g (m)
f
f ).
(35)
m
(b)
FIGURE 11. Examples of the application of (a) PPB filter and (b) SAR–BM3D filter to the 4-look COSMO-SkyMed image in Fig. 10-(a).
SEPTEMBER 2013
23
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
In the above equation, the weights w (n, m) can be thought of
as a measure indicating to what extent a pixel at position m
has the same distribution as the reference pixel at position n.
The definition of the weights w (n, m) is the key problem
of the WMLE approach. In the PPB filter, this problem is
solved by expressing the weights as the probability, given
the observed image g, that two patches centered at positions n and m can be modeled by the same distribution. By
assuming the independence of the pixels of the patches, the
weights can be formally expressed as
w (n, m) = % p (f (n + k) = f (m + k) g (n + k),
k
g (m + k)) 1/h,
(36)
where k varies over the image patch and h is a decay parameter. According to a Bayesian framework, without knowledge of the prior probabilities p (f (n + k) = f (m + k)),
the posterior probabilities in equation (36) can be
assumed proportional to the likelihood p (g (n + k),
g (m + k) f (n + k) = f (m + k)). This permits to adapt the
weights of the PPB filter to several image distributions. For
the case of SAR images, by assuming that pixel amplitudes
a = g are modeled as independent and identically distributed according to a Nakagami-Rayleigh distribution,
the PPB weights can be derived as [108]
a (n + k)
a (m + k)
1
w (n, m) = exp =- h / log d a (m + k) + a (n + k) nG (37)
k
and the despeckled image can be obtained according to the
WMLE as
2
tf (n) = / m w (n, m) a (m) .
/ m w (n, m)
(38)
In [108], the model is further improved by letting the
probabilities in (36) depend also on a previous estimate
of the noise-free image. This leads to an iterative filtering
approach, in which the weights are updated at each iteration according to the previous result of the filter. For the
detailed derivation of the iterative PPB filter, the interested
reader is referred to [108].
An example of the application of the PPB filter to the
COSMO-SkyMed image in Fig. 10-(a) is given in Fig. 11-(a).
As it appears, PPB filtering overly smooths textures, if any,
and tries to achieve a hard segmentation of the scene also
in the presence of softly switching classes.
B. SAR BLOCK MATCHING 3-D FILTER
The SAR block matching 3-D (SAR-BM3D) filter, proposed
by Parrilli et al. in [109], is a SAR-oriented version of the
block matching 3-D filter [111], which applies the NL principle in combination with a wavelet representation. The
key idea of the BM3D filter is to apply the NL principle for
collecting groups of similar image patches, and to compute
24
a wavelet decomposition of the resulting 3-D blocks. The
NL grouping of similar patches is expected to form a highly
correlated 3-D signal, which will likely have a very sparse
representation in the wavelet domain, leading to an effective separation of noise–free and noisy coefficients.
The processing flow of the BM3D filter can be summarized by the following steps: 1) for each reference patch in
the observed image, collect the most similar patches according to a Euclidean distance criterion, and form a 3-D group;
2) apply 3-D wavelet transform, denoising of wavelet coefficients, and inverse transform; 3) return all filtered patches
to their original positions, and combine them using suitable
weights. It is worth noting that the above approach can be
seen as a collaborative filtering: in general, the patches will
be highly overlapped, so that each filtered pixel will result
from the combination of several filtered patches. In [111],
the final BM3D filter is obtained by repeating the above
processing flow in a two iteration procedure. In the first
step, wavelet domain denoising is achieved by simple hardthresholding, in order to yield a coarsely denoised image.
The second step uses the denoised image obtained after
the first step to improve the 3-D grouping accuracy, and
replaces hard-thresholding with Wiener filtering, where the
energy spectrum of the noise-free image is estimated form
the coarsely denoised image.
In order to adapt the BM3D filter to the case of SAR
images, the SAR-BM3D filter considers two main modifications: 1) the similarity measure between patches is computed according to (37), following the same approach as in
[108]; hard-thresholding and Wiener filtering are replaced
with an LMMSE estimator [78] based on the additive signal-dependent noise model in (25). According to the two
step procedure of the original BM3D filter, in the first step
the LMMSE estimator is based only on the observed image
g and the filtered wavelet coefficients are obtained according to equation (27), whereas in the second step it approximates the moments of the noise-free image according to the
output of the first step, and the filtered wavelet coefficients
are obtained according to
t f (n) =
W
t 2f,1 (n)
W
t (n) + v 2Wv W g (n),
W
(39)
2
f, 1
t f,1 are the noise-free wavelet coefficients estimated
where W
at the first step and the variance of the wavelet coefficients
of the signal-dependent noise is obtained as
2
vW
v =
1 /7
t f,1 (k)A2
W g (k) - W
|G | k!G
(40)
with G denoting the set of wavelet coefficients belonging
to a 3-D group. The final despeckled image is obtained as
a weighted average of the overlapped denoised patches,
where the weights for each patch are inversely proportional
to the corresponding value of v 2Wv [109].
An example of the application of the SAR-BM3D filter to
the COSMO-SkyMed image in Fig. 10-(a) is given in Fig. 11-(b).
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
VII. TOTAL VARIATION REGULARIZATION
Image denoising through TV regularization can be defined
as the solution of a minimization problem
tf = arg min J (f, g),
(41)
f
where the cost function to be optimized can be expressed as
J (f, g) = U (f) + mW (f, g) .
(42)
In the above equation, U (f) denotes a regularization term
including prior information about the noise-free image f,
whereas W (f, g) denotes a data fidelity term.
The regularization term is usually defined as the TV
norm of the noise-free image, i.e.
U (f) = / n df (n) ,
(43)
where df (n) denotes the magnitude of the gradient of f
and can be computed as
df (n) = fx (n) 2 + fy (n) 2 ,
(44)
where fx (n) and fy (n) denote horizontal and vertical
first order differences evaluated at pixel n, respectively.
The minimization of the TV norm tends to promote a
piecewise smooth image, which is usually a good prior
for natural images, since it preserves important structures
like edges.
The data fidelity term can be defined according to several different approaches. A popular approach is to set the
data fidelity term equal to the negative of the log-likelihood
of f given the observed image g, that is
W ( f, g) = - log p ( f g) .
(45)
If the TV norm is interpreted as a negative log-prior term,
i.e., U ( f) = - log p ( f ) it is evident that the solution of the
problem in (41) is equivalent to the MAP estimate of f.
A. DESPECKLING USING TV REGULARIZATION
When it comes to despeckling, the main problem is adapting the TV framework to the multiplicative noise model. In
[119], the above problem is tackled in the original intensity
domain by assuming a Gamma-distributed speckle, which
in turn implies a Gamma p ( f g). The resulting problem
can be expressed as
tf = arg min /
f
n
df (n) + m / d log f (n) +
n
g (n)
n.
f (n)
(46)
Despite its elegance, the above approach suffers from the
fact that the functional is convex only for 0 < f < 2g. In
order to obtain a convex problem, several authors have
considered the logarithmic domain. A simple solution is to
keep the same data fidelity term as in (46), but to replace f
by f l = log f in the regularization term [120], which yields
tf l = arg min /
fl
SEPTEMBER 2013
n
df l (n) + m / _ f l (n) + g (n) e -f l (n) i . (47)
n
FIGURE 12. Example of the application of the TV filter in [122] to the
4-look COSMO-SkyMed image in Fig. 10-(a).
Another natural solution is to consider a quadratic data
fidelity term [117], [120], which yields
tf l = arg min /
fl
n
df l (n) + m / ^ f l (n) - g l (n) h .
2
(48)
n
Interestingly, the above solution is still equivalent to a MAP
estimate when we can approximate the logarithmically
transformed speckle as Gaussian.
All of the above approaches consider solutions in
the spatial domain. In [122], Durand et al. propose to
combine the advantages of the TV regularization framework with those offered by a sparse representation. The
proposed solution consists in computing the data fidelity term in the domain of a redundant multiscale representation. The rationale is that relevant structures in
the image are more effectively preserved in a multiscale
representation, while a TV prior helps in removing characteristic artifacts caused by wavelet thresholding. In
order to limit the effects of noisy wavelet coefficients,
the authors of [122] propose to compute the data fidelity term on a hard-thresholded version of the observed
coefficients, where the coefficients are obtained by applying a curvelet transform to the log-transformed intensity
image. The authors also suggest using an under optimal
threshold, so as to preserve as much as possible curvelet coefficients relevant to edges and textures. In order to
take into account the long tailed distribution of curvelet
coefficients, the data fidelity term is defined as the mean
absolute error between the despeckled coefficients and
the hard-thresholded coefficients. The final optimization
problem can be expressed as
25
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
tf l = arg min f l / n
df l (n)
+ m / n W f l (n) - H [W gl (n)]
(49)
where H [$] denotes the hard-thresholding operator.
Since the above estimator is prone to bias, the authors
of [122] propose to compute the despeckled image as
tf = exp (tf ') (1 + z 1 (L) /2), where z 1 (L) is the first-order
polygamma function and represents the variance of L-look
log-transformed speckle [33].
In general, the solution of the aforementioned optimization problems requires a suitable minimization scheme.
According to the properties of the functional to be minimized, several schemes can be used, including gradient projection [118], iterative splitting methods [122], [123], inverse
scale space flow [120]. The details of such minimization
schemes are beyond the scope of this tutorial and the related
literature is really vast. The interested reader is referred to
the above cited papers and the references therein.
An example of the application of the filter proposed in
[122] to the COSMO-SkyMed image in Fig. 10-(a) is given
in Fig. 12.
VIII. ASSESSMENT OF DESPECKLING FILTERS
One of the most challenging tasks is the validation and
quality assessment of data processed for speckle reduction.
The most evident problem is that the noise-free reflectivity that we wish to estimate is unknown, so that no
comparison can be carried out between the output of the
despeckling process and the actual ground truth. Another
important issue is the relationship between quality and
fidelity of despeckled SAR data. Like many other denoising
frameworks, the quality of a processed SAR image is usually evaluated in terms of blurring of homogeneous areas,
i.e., suppression of speckle noise, and detail preservation
in heterogeneous areas. Nonetheless, in incoherent SAR
imagery, a fundamental part of the information is represented by the relative values of the reflectivity of the targets, which allow measurements and inferences on the target scene. Consequently, the radiometric preservation of
the signal is an important requirement: a good despeckling
filter should not introduce bias on the reflectivity.
An immediate and subjective approach for quality
assessment is represented by visual inspection of filtered
images. Visual inspection permits detection of the main
human–visible features that characterize the behavior of a
despeckling filter. Such features include edge preservation
capability, degree of blur, point target preservation, as well
as structural artifacts which are hardly detected by objective and direct measurements. On the other hand, visual
assessment does not allow either quantitative comparisons
between the performances of different despeckling filters
to be made or the bias introduced by the filter to be effectively estimated.
In order to overcome the limitations of visual comparison, several objective performance indexes have been
26
proposed in the literature for the quality assessment of
despeckling filters. They can be mainly divided into two
classes: with–reference and without–reference indexes.
With–reference indexes are commonly used in the
image denoising field. Their use implies that the noise–
free, or reference, image is known. A typical approach consists in choosing a reference image, either optical or synthetic, representing the actual reflectivity or ground–truth,
and creating a synthetically degraded version according to
a given signal model. These indexes permit a quantitative
and objective comparison between the performances of
different filters, which are expected to perform similarly
on real SAR images. Moreover, insights on filters behavior on specific image features, like edge preservation and
homogeneous areas smoothing, can be easily highlighted
by choosing appropriate reference images and even synthetic–generated patterns. Unfortunately, experimental
results carried out on simulated SAR images often are not
sufficient to infer the performances of despeckling filters
on real SAR images, since the synthetically speckled image
may not be consistent with the actual SAR image formation and acquisition processes. Furthermore, the statistical properties of the chosen reference image and of a real
ground–truth reflectivity can substantially differ.
On the contrary, without–reference indexes do not trust
on the knowledge of the ground–truth. They are uniquely
based on specific statistical hypotheses on the signal model.
Since the signal model is strongly dependent on the degree
of scene heterogeneity, a supervised selection of the most
appropriate areas for the computation of a specific index,
e.g., homogeneous areas, may be required.
In the following, the most used indexes belonging to
both the above mentioned classes are presented. Note that
the statistical operator of expectation E [$] and the moments
of the involved variables, such as the variance and covariance, here denoted as Var[$] and Cov [$] for the sake of simplicity, should be replaced by their empirical versions based
on spatial averages when evaluating the indexes.
A. WITH–REFERENCE INDEXES
The mean square error (MSE), or Euclidean distance, between
the ground–truth f and the despeckled image tf, and other
measures derived from the MSE, like the signal–to–noise
ratio (SNR), the peak signal–to–noise ratio (PSNR) and the
energy signal-to-noise ratio (ESNR), have been widely used
for the quality assessment of both denoising and despeckling [33], [57], [76]. Unlike the case of additive signal–independent noise, in the presence of signal–dependent noise
the MSE is strongly influenced by the average signal level of
the ground truth. Consequently, a quantitative evaluation
of despeckling filters using this kind of indexes is strongly
dependent on the content of the ground–truth image, even
though performance hierarchy is usually preserved across
different images.
MSE–based measurements are useful to obtain a global
performance assessment on the whole image, but usually
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
TABLE 1. LIST OF COMMONLY USED WITH-REFERENCE INDEXES FOR EVALUATING PERFORMANCES
OF DESPECKLING ALGORITHMS.
INDEX
NOTE
t
t
MSE = E [(f - f ) 2]
f, f : speckle–free and despeckled images
Var [f ]
SNR = 10 $ log 10 ; MSE E
Var [f] : speckle–free image variance
2
PEAK
f
PSNR = 10 $ log 10 ; MSE E
fPEAK : maximum value allowed by the samples dynamic range
2
E [f ]
ESNR = 10 $ log 10 ; MSE E
E [f 2] : speckle–free image power
t
t
2 $ E 6fp@ $ E 7fpA + C 1
1
MSSIM = M / Mp =-01 >
E 6f 2p@ + E 7f 2pA + C 1
t
EC =
Cov 7f H, f HA
Var 6f H@ $ Var 7f HA
FOM =
1
1
/N
2
max (N, N) n = 1 1 + d n a
t
t
2 $ Cov 7fp, fpA + C 2
H
Var 6fp@ + Var 7fpA + C 2
t
t
t
they yields little information about the preservation of
specific features, for which other indexes can be used.
The mean structural similarity index measurement (MSSIM)
[135], proposed for the general denoising framework and
adopted also in the context of despeckling, underlines
the perceived changes in structural information after the
filtering process. MSSIM takes values over the interval [0,
1], where 0 and 1 indicate no structural similarity and
perfect similarity, respectively. As demonstrated in [135],
MSSIM can substantially differ between images having
very similar MSE values.
The edge correlation (EC) index has been proposed as
a measure of edge preservation for despeckling of echographic images [136] and has been extended to the SAR field
[72]; it is defined as the correlation coefficient (0 # EC # 1)
between high pass versions of the original and despeckled
images. This index may be distorted by possible residual
speckle noise that is enhanced by the high pass filtering.
Another index of edge preservation is Pratt’s figure of
merit (FOM), which has been used in [96] for the quality
assessment of despeckled SAR and ultrasound images.
FOM is defined on a local patch of the image containing
an edge. The more similar the edge maps, the closer to zero
the FOM values. Consequently, this index is strictly related
to the map edge detector that is used, which is crucial especially for the despeckled image when a residual noise component is present.
Table 1 summarizes the above mentioned indexes.
A synthetically speckled images has been produced
starting from a 512 # 512 digitized aerial photograph of
San Francisco. The original speckle-free image, regarded
as an amplitude format, has been squared and multiplied
by an exponentially distributed fading term, in order to
simulate a single-look SAR image in intensity format.
The simulated speckle is spatially uncorrelated and fully
SEPTEMBER 2013
t
fp, fp, p = 0, f, M - 1: speckle–free and despeckled image patches;
C 1, C 2: suitable constants.
t
f H, f H: high pass–filtered speckle–free and despeckled images
t
N, N: number of points belonging to an edge in speckle–free and
despeckled image patches; d 2n: Euclidean distance between the edge
pixels in the despeckled image patch and the nearest ideal edge
pixel in the speckle–free one; a: suitable constant.
developed. The noisy intensity image, together with all
filtered intensity versions, has been square rooted, for
displaying convenience, and is shown together with the
8-bit original, regarded as an amplitude image, in Fig.
13-(b) and Fig. 13-(a), respectively.
The filters compared in this subsection are representative of different approaches to despeckling described in
this paper: Kuan [52] and C–MAP [54] as classical spatial
filters; GG–MAP–S [83] and LG–MAP–C [85] as Bayesian filters in the wavelet domain (input format is square
root of intensity [137]); Probability Patch–Based (PPB)
[108] and SAR–BM3D [109] as non–local mean filters
in the spatial and wavelet domain; L1 Fidelity on Frame
Coefficients (L1–FFC) [122] as a TV-based filter. Visual
comparisons of the results obtained with the same filters
can be made observing Fig. 13. What immediately stands
out is that local spatial filter (Kuan and C–MAP) are
unable to clean the noisy background. A residual inhomogeneity, like a coarse granular texture, is noticeable
especially on the sea. This effect is thoroughly missing
in wavelet-domain filters, as well as in nonlocal-mean
and TV filters. Preprocessing of point targets was disabled
in wavelet schemes, because the simulated speckle is
fully developed.
Fig. 14 shows the performance indexes obtained by
means of the test despeckling filters.
B. WITHOUT–REFERENCE INDEXES
As previously stated, without–reference indexes do not rely
on the complete knowledge of the true reflectivity, but are
based on the statistical model of the SAR signal as well as
on some simple assumptions on the degree of heterogeneity of the underlying scene.
The equivalent number of look (ENL) [46] is an index
suitable for evaluating the level of smoothing in
27
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
(i)
FIGURE 13. Results on a synthetically speckled image: (a) noise-free reference, (b) noisy (1-look), (c) Kuan, (d) C-MAP, (e) GG–MAP–S,
(f) LG–MAP–C, (g) PPB, (h) SAR–BM3D, and (i) L1–FFC.
homogeneous areas, that is where the scene variation
is supposed to be negligible with respect to speckle
noise fluctuations. The ENL of the original SAR image
is related to the nominal number of looks through the
autocorrelation function of speckle [142], whereas it
increases after the despeckling stage according to the
smoothing capability of the filter.
Other typical measures can be computed from the ratio
image r, defined as the point–by–point ratio between the
noisy and the filtered image [4]
28
g (n)
(50)
tf (n) .
The ratio image is a useful information in both homogeneous and heterogeneous scenes, wherever the fully developed speckle model holds. It represents the noise pattern
removed by the despeckling filter that, according to the
model, should be C-distributed. An ideal filter should
result in a pure random pattern, whereas poor speckle noise
removal causes structural information, such as borders and
edges, to be clearly visible in the ratio image. The mean and
r (n) =
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
30
25
20
15
10
5
0
PSNR (dB)
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
EC
MSSIM
Noisy
LG-MAP-C
Kuan
PPB
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
Gamma-MAP
SAR-BM3D
FOM
GG-MAP-S
L1-FFC
FIGURE 14. With-reference indexes computed for the image in Fig. 13-(b), with the image Fig. 13-(a) as speckle–free reflectivity.
the variance of r, that is n r = E [r] and v 2r = Var[r], should
be as close as possible to the respective theoretical statistical moments of the speckle noise process. For this reason,
they are often used as indexes of bias and speckle power
suppression, respectively. A measure of bias is also given by
the B index [30], in which a value close to zero indicates an
unbiased estimation.
Under the hypothesis of multiplicative speckle noise, a
measure of texture preservation on heterogeneous areas is
given by the comparison between the coefficient of variation
calculated on the despeckled image, namely Ctf , and its the
expected theoretical value on the noise-free image, C f [20].
Intuitively, a poor preservation of details yields C f > Ctf ,
while the introduction of impairments leads to C f < Ctf .
Since the speckle model does not hold in the presence
of persistent scatterers or point targets, despeckling filters
should keep their values unchanged. A point target is usually characterized by a cluster of pixels whose reflectivity
values are much higher, even some orders of magnitude,
than the mean reflectivity of the surrounding scene. The
target–to–clutter ratio (TCR) [138], [139] aims at measuring
the relative value of strong scatterers with respect to the values of the surrounding pixels. TCR values computed before
and after despeckling are indicative about how much a filter preserves the radiometric properties in the patch.
Table 2 summarizes the most commonly used withoutreference indexes for evaluating despeckling algorithms
performance. Fig. 15 shows the without–reference indexes
obtained on the image in Fig. 10-(a). The indexes have been
computed for the original 1024 # 1024 image and for a
512 # 512 4–look version, generated by means of spatial
multilooking (2 # 2 average).
C. DISCUSSION
The computational complexities of the most relevant filters among those reviewed raises an interesting concern.
Early spatial filters are nowadays real-time (less than 1 s
to process a 1024 # 1024 scene on a standard platform).
Wavelet-based methods are at least ten times longer to run,
up to one hundred times for GG–MAP–S, which requires
numerical calculation of the maximum of a function [83],
unlike LMMSE and LG–MAP–C, which admit closed form
SEPTEMBER 2013
TABLE 2. LIST OF COMMONLY USED WITHOUTREFERENCE INDEXES FOR EVALUATING PERFORMANCES
OF DESPECKLING ALGORITHMS.
INDEX
ENL =
NOTE
t
t
2
E7f A
Var 7 f A
f, f : speckle–free and despeckled
images; ENL is evaluated in homogeneous areas
t
n r = E [r], v 2r = Var [r]
B = E=
Cf =
t
r (n ) =
g (n )
tf (n) : ratio image
t
(g - f ) G
g
t
Var 7 f A
E7f A
t
TCR = 20 log 10
C 2g - C 2u
(expected value);
1 + C 2u
C g, , C u: coefficients of variation of
the observed noisy image g and
of the speckle noise u
Cf =
max P 6g@
E P 6g@
P: patch containing a point
target; max P , E P computed over
the patch
solutions. For all multiresolution methods, biorthogonal
9/7 wavelet filters and four levels of decomposition (corresponding to a baseband approximation having 4 4 = 256
nominal looks) have been used. Biorthogonal filters are
preferred to orthogonal filters in image processing applications because they allow filters of different lengths, and
hence of spectral selectivity, to be available for the low pass
(9 coeffs.) and high pass (7 coeffs.) analyses. Conversely,
NL filtering approaches, in either space (PPB) or wavelet
(SAR–BM3D) domain, have a significantly higher computational cost, mainly because of iterated processing, with
recalculation of statistics after each step. Eventually, the
TV-based filter examined (L1–FFC) is comparable with NL
filters. The numbers of iterations are those recommended
by the respective authors in their implementations. Table 3
summarizes the complexity of despeckling algorithms.
A key point in the despeckling of SAR images is the
extent to which models assumed for the signal or the
noise match the actual statistics of the data. By observing the results on 1-look data in Fig. 15, it is quite evident
that all filters yield a biased outcome ^n r < 1, B > 1 h and
a limited speckle removal capability ^v 2r < 1 h . Both these
29
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
100
1
80
0.95
60
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.9
40
0.85
20
0
0.8
ENL
Mean[r]
Noisy
LG-MAP-C
Var[r]
Gamma-MAP GG-MAP-S
SAR-BM3D
L1-FFC
Kuan
PPB
1.2
1
0.8
0.6
0.4
0.2
0
Coeff_Var CF
(a)
250
1.1
0.4
200
1.05
0.3
150
1
100
0.95
50
0.9
0
0.2
0.1
0
0.85
ENL
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Mean[r]
Noisy
LG-MAP-C
Var[r]
Kuan
PPB
Gamma-MAP
SAR-BM3D
Coeff_Var CF
GG-MAP-S
L1-FFC
(b)
FIGURE 15. Without-reference indexes computed for different despeckling algorithms for the image in Fig. 10-(a). (a) One-look image (theoretical value C f = 0.355 ); (b) 4-look version of the same image (theoretical value C f = 0.544 ).
effects occur because all filters do not take into account that
speckle is spatially autocorrelated in real single–look SAR
images [30], for the following reasons: 1) oversampling of
SAR raw data with respect to the Nyquist rate given by twice
the chirp bandwidth; 2) frequency windowing applied
when the raw data are focused and aimed at improving the
response of targets, avoiding Gibbs’ effects. As reported in
Fig. 15 for 4-looks data, multilooking allows all filters to
obtain values of n r and v 2r closer to the ideal ones. This
mainly happens because the multilooking process reduces
the speckle correlation; unfortunately, it also halves the
image resolution in both range and azimuth directions.
Very few despeckling filters that specifically consider
the speckle correlation have been proposed in the literature
(e.g., [30]). Recently, a blind speckle decorrelation method
to be applied to SLC images has been proposed [31], [141]
to enhance the performances of existing despeckling filters.
The idea is estimating the SAR system frequency response
on the original SLC image in order to compensate its effect
by an inverse filtering (whitening stage), so that an SLC image
having uncorrelated speckle noise, but preserving the radiometric features, is produced. In [140], it has been shown
that the introduction of the whitening stage allows noticeable performance gains for filters based on the uncorrelated
speckle model. A visual and numerical example on a single–
look COSMO–SkyMed image is proposed in Fig. 16. The correlation coefficient t dramatically decreases after the whitening stage. MAP–GG–S outperforms its own results when
it is applied to the uncorrelated speckled image. The problem
of speckle correlation occurs only for one-look data, because
the process of multilooking, equivalent to low pass filtering
and decimation, lowers the correlation coefficient (CC) of
speckle from about 35% to less than 10% [142].
A visual analysis of the image details in Fig. 16 highlights
that wavelet despeckling suffers from the presence of structured artifacts mainly located around edges, referred to as
glitches, that are due, in order of importance, to: 1) speckle
correlation, 2) input image format (amplitude is preferable
to intensity, because yields a more accurate MAP estimation
in UDWT domain [143]), 3) type of wavelet filter (the shortest filters of Haar transform [89] produce the least noticeable
artifacts). Also the type of decomposition (à trous wavelet
(ATWT) [37] rather than UDWT) is a topic worth being
investigated, also because ATWT accommodates all details
TABLE 3. COMPUTATIONAL COMPLEXITY OF DESPECKLING METHODS. BETWEEN TWO CONSECUTIVE GRADES THERE IS
APPROXIMATELY ONE ORDER OF MAGNITUDE. SO, PPB IS ABOUT 1000 TIMES SLOWER THAN KUAN’S FILTER, ON THE
SAME COMPUTING PLATFORM.
30
Filter
Kuan (7
Complexity
Very low
#
7)
C –MAP (7
Very low
#
7)
UDWT LMMSE
LG–MAP–C
GG–MAP–S
PPB
SAR–BM3D
L1–FFC
Low
Low
Medium
High
High
High
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
(a)
(b)
(c)
(d)
FIGURE 16. (a) Original one-look COSMO-SkyMed image ( t - 0.29 ); (b) MAP–GG–S [83] of original (ENL = 27.90, n r = 0.940, v 2r = 0.702 );
(c) whitened [140] ( t - 0.05 ); (d) MAP–GG–S of whitened (ENL = 142.29, n r = 0.997, v 2r = 0.936 ). t is the CC of speckle measured on the
complex image [29]; ENL, n r and v 2r are calculated on a homogeneous patch after despeckling.
of one scale in a unique plane; thus the number of coefficients to be despeckled, and hence computing times, would
be three times lower. However, the adaptivity with orientation featured by UDWT would be lost with ATWT.
The assessment of the performances of despeckling filters on real SAR data is often problematic due to the lack
of with–reference indexes. In order to overcome such problems, a possible idea is to use electromagnetic SAR image
generators [23]. Such simulators are based on more physical–oriented models, which consider the propagation of the
electromagnetic wave and its interaction with targets and
surfaces, and usually require a more detailed parametric
description of the target scene with respect to the models
used in signal processing applications. In [22], the authors
use an electromagnetic SAR image generator to simulate
several independent acquisitions of the same scene. If the
number of acquisitions is sufficiently high, their average
can be considered as a good approximation of the noise–
free reflectivity and can be used to compute with–reference
indexes. The advantage of this technique is that the simulated images do not necessarily obey the fully developed
speckle model and provide insights on the behavior of the
filter on point targets and highly heterogeneous areas. On
the other hand, the underlying reflectivity follows a synthetically generated pattern, which may not be fully representative of the reflectivity usually encountered in real SAR
images, especially in complex scenes, due to the ideal models of objects fed to the simulator.
Another viable approach to devise a fully automatic
method for quality assessment of despeckled SAR images
was recently proposed by the authors [144]. The rationale of
the new approach is that any structural perturbation introduced by despeckling, e.g., a local bias of mean or the blur
of a sharp edge or the suppression of a point target, may
be regarded either as the introduction of a new structure
or as the suppression of an existing one. Conversely, plain
removal of random noise does not change structures in the
image. Structures are identified as clusters in the normalized
scatterplot of original to filtered image. Ideal filtering should
produce clusters all aligned along the main diagonal. In pracSEPTEMBER 2013
tice clusters are moved far from the diagonal. Cluster centers
are detected through the mean shift algorithm. A structural
change feature is defined at each pixel from the position and
population of off-diagonal clusters [145]. Such a feature may
be regarded as a spatial map of filtering inaccuracies. A preliminary validation has been carried out on simulated SAR
images, with a good correlation between feature and objective filtering error. In experiments on COSMO-SkyMed
images, the automatic ranking of filters matches the subjective trials of experts. The proposed feature detects filtering
impairments but is unable to measure the overall effectiveness of filtering. Therefore, its use must be coupled with
another index measuring the effectiveness of cleaning, e.g.,
ENL, regardless of its accuracy.
IX. CONCLUSIONS AND PERSPECTIVES
This tutorial has demonstrated that despeckling of SAR
images takes into account several issues related to signal and
noise modeling, signal representation, estimation theory and
quality assessment. Concerning Bayesian estimation, starting from Lee filter, local-window adaptive filtering has been
progressively enhanced, up to a saturation of performances,
due to the trade off of using windows small enough to retain
edges textures and fine details and large enough to allow a
consistent and confident statistical estimate to be achieved.
In the last two decades, the introduction of multiresolution analysis has been found to boost despeckling
algorithms performances. Key points of wavelet-based
despeckling is the modeling of the reflectivity and of the signal-dependent noise in the wavelet domain and the choice
of the estimation criterion to achieve the noise-free wavelet
coefficients. While several authors have chosen overfitting
models sacrificing space adaptivity, others have tried to
keep the advantages of an adaptivity in both scale and space
by using pdfs with few parameters to be estimated locally on
subbands/frames. A preprocessing step of point targets that
must retain their radiometry after despeckling and a segmented approach, in which sample statistics are calculated
on homogeneous segments, complete Bayesian despeckling
in wavelet domain.
31
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
As to non-Bayesian approaches, Lee’s sigma filter has
evolved into bilateral filtering, which has possibly inspired
nonlocal filtering, excellent examples of which are found
both in spatial and in wavelet domains. In parallel, total
variation has emerged as a powerful regularization technique that can be specialized to the signal dependent noise
model and allows constraints to be set on several mathematical properties of the output image. As an example,
setting a constraint on L1 norm definitely avoids glitches
and other impairments. Presently, computational issues
are mainly responsible for the moderate, yet increasing,
popularity of such methods among users.
New horizons are undoubtedly in the direction of compressed sensing, of which denoising seems to be one of the
most promising application, notwithstanding objective
difficulties come from the signal dependent, and hence
nonstationary, noise model. Given the huge effort of
researchers in this area, new developments and applications to despeckling are expected in a near future. Computational issues are also the main drawback of algorithms
based on compressed sensing, with respect to spatial and
wavelet domain Bayesian algorithms. The ever increasing
diffusion of multiprocessor systems will be beneficial for
methods that can be easily parallelized.
X. ACKNOWLEDGMENTS
The authors are grateful to C. Deledalle and to L. Verdoliva
for providing us the codes of PPB and SAR-BM3D methods,
respectively. The COSMO-SkyMed data were kindly made
available by the Italian Space Agency (ASI).
REFERENCES
[1] A. Moreira, P. Prats-Iraola, M. Younis, G. Krieger, I. Hajnsek,
and K. P. Papathanassiou, “A tutorial on synthetic aperture radar,” IEEE Geosci. Remote Sensing Mag., vol. 1, no. 1, pp. 6–43,
Mar. 2013.
[2] L. Alparone, S. Baronti, A. Garzelli, and F. Nencini, “Landsat
ETM+ and SAR image fusion based on generalized intensity
modulation,” IEEE Trans. Geosci. Remote Sensing, vol. 42, no. 12,
pp. 2832–2839, Dec. 2004.
[3] F. Argenti, G. Torricelli, and L. Alparone, “MMSE filtering of
generalised signal-dependent noise in spatial and shift-invariant
wavelet domains,” Signal Process., vol. 86, no. 8, pp. 2056–2066,
Aug. 2006.
[4] C. Oliver and S. Quegan, Understanding Synthetic Aperture Radar
Images. Boston, MA: Artech House, 1998.
[5] L. M. Novak and M. C. Burl, “Optimal speckle reduction in polarimetric SAR imagery,” IEEE Trans. Aerosp. Electron. Syst., vol.
26, no. 2, pp. 293–305, Mar. 1990.
[6] J.-S. Lee, M. R. Grunes, and S. A. Mango, “Speckle reduction in
multipolarization, multifrequency SAR imagery,” IEEE Trans.
Geosci. Remote Sensing, vol. 29, no. 4, pp. 535–544, July 1991.
[7] R. Touzi and A. Lopès, “Principle of speckle filtering in polarimetric SAR imagery,” IEEE Trans. Geosci. Remote Sensing, vol. 32,
no. 5, pp. 1110–1114, Sept. 1994.
[8] J.-S. Lee, M. R. Grunes, and G. De Grandi, “Polarimetric SAR
speckle filtering and its implication for classification,” IEEE
Trans. Geosci. Remote Sensing, vol. 37, no. 5, pp. 2363–2373,
Sept. 1999.
[9] J.-S. Lee, M. R. Grunes, D. L. Schuler, E. Pottier, and L. FerroFamil, “Scattering-model-based speckle filtering of polarimetric
32
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
SAR data,” IEEE Trans. Geosci. Remote Sensing, vol. 44, no. 1, pp.
176–187, Jan. 2006.
R. J. Dekker, “Speckle filtering in satellite SAR change detection
images,” Int. J. Remote Sens., vol. 19, no. 6, pp. 1133–1146, Apr.
1998.
S. Quegan and J. J. Yu, “Filtering of multichannel SAR images,”
IEEE Trans. Geosci. Remote Sensing, vol. 39, no. 11, pp. 2373–
2379, Nov. 2001.
J. Bruniquel and A. Lopès, “Multivariate optimal speckle reduction in SAR imagery,” Int. J. Remote Sens., vol. 18, no. 3, pp.
603–627, Feb. 1997.
R. Touzi, A. Lopès, J. Bruniquel, and P. W. Vachon, “Coherence
estimation for SAR imagery,” IEEE Trans. Geosci. Remote Sensing,
vol. 37, no. 1, pp. 135–149, Jan. 1999.
P. A. Rosen, S. Hensley, I. R. Joughin, F. K. Li, S. N. Madsen, E.
Rodriguez, and R. M. Goldstein, “Synthetic aperture radar interferometry,” Proc. IEEE, vol. 88, no. 3, pp. 333–382, Mar. 2000.
J.-S. Lee, K. P. Papathanassiou, T. L. Ainsworth, M. R. Grunes,
and A. Reigber, “A new technique for noise filtering of SAR interferometric phase images,” IEEE Trans. Geosci. Remote Sensing,
vol. 36, no. 5, pp. 1456–1465, Sept. 1998.
B. Aiazzi, S. Baronti, M. Bianchini, A. Mori, and L. Alparone,
“Filtering of interferometric SAR phase images as a fuzzy
matching-pursuit blind estimation,” EURASIP J. Appl. Signal
Process., vol. 2005, no. 20, pp. 3220–3230, 2005.
M. Costantini, “A novel phase unwrapping method based on
network programming,” IEEE Trans. Geosci. Remote Sensing, vol.
36, no. 3, pp. 813–821, May 1998.
B. Aiazzi, L. Alparone, S. Baronti, and A. Garzelli, “Coherence
estimation from incoherent multilook SAR imagery,” IEEE
Trans. Geosci. Remote Sensing, vol. 41, no. 11, pp. 2531–2539,
Nov. 2003.
J.-S. Lee, I. Jurkevich, P. Dewaele, P. Wambacq, and A. Oosterlinck, “Speckle filtering of synthetic aperture radar images: A
review,” Remote Sens. Rev., vol. 8, no. 4, pp. 313–340, 1994.
R. Touzi, “A review of speckle filtering in the context of estimation theory,” IEEE Trans. Geosci. Remote Sensing, vol. 40, no. 11,
pp. 2392–2404, Nov. 2002.
J.-S. Lee, “Digital image enhancement and noise filtering by
use of local statistics,” IEEE Trans. Pattern Anal. Mach. Intell., vol.
PAMI-2, no. 2, pp. 165–168, Feb. 1980.
G. Di Martino, M. Poderico, G. Poggi, D. Riccio, and L. Verdoliva, “Benchmarking framework for SAR despeckling,” IEEE
Trans. Geosci. Remote Sensing, to be published.
G. Franceschetti, M. Migliaccio, D. Riccio, and G. Schirinzi,
“SARAS: A synthetic aperture radar SAR raw signal simulator,”
IEEE Trans. Geosci. Remote Sensing, vol. 30, no. 1, pp. 110–123,
Jan. 1992.
C. Elachi and J. J. van Zyl, Introduction to the Physics and Techniques of Remote Sensing. Hoboken, NJ: Wiley, 2006.
F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave Remote Sensing:
Active and Passive, Volume II: Radar Remote Sensing and Surface Scattering and Emission Theory. Norwood, MA: Artech House, 1986.
J. W. Goodman, “Some fundamental properties of speckle,” J.
Opt. Soc. Amer., vol. 66, no. 11, pp. 1145–1150, Nov. 1976.
M. Tur, K. C. Chin, and J. W. Goodman, “When is speckle multiplicative?” Appl. Optics, vol. 21, no. 7, pp. 1157–1159, Apr. 1982.
J. C. Curlander and R. N. McDonough, Synthetic Aperture Radar:
Systems & Signal Processing. New York: Wiley, 1991.
S. N. Madsen, “Spectral properties of homogeneous and nonhomogeneous radar images,” IEEE Trans. Aerosp. Electron. Syst.,
vol. AES-23, no. 4, pp. 583–588, July 1987.
S. Solbø and T. Eltoft, “A stationary wavelet-domain Wiener filter for correlated speckle,” IEEE Trans. Geosci. Remote Sensing,
vol. 46, no. 4, pp. 1219–1230, Apr. 2008.
A. Lapini, T. Bianchi, F. Argenti, and L. Alparone, “A whitening
method for the despeckling of SAR images affected by correlated speckle noise,” in Proc. 20th European Signal Processing Conf.
(EUSIPCO), 2012, pp. 2487–2491.
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
[32] H. H. Arsenault and M. Levesque, “Combined homomorphic
and local-statistics processing for restoration of images degraded
by signal dependent noise,” Appl. Opt., vol. 23, no. 6, pp. 845–
850, Mar. 1984.
[33] H. Xie, L. E. Pierce, and F. T. Ulaby, “Statistical properties of
logarithmically transformed speckle,” IEEE Trans. Geosci. Remote Sensing, vol. 40, no. 3, pp. 721–727, Mar. 2002.
[34] I. Daubechies, “Orthonormal bases of compactly supported wavelets,” Commun. Pure Appl. Math., vol. 41, no. 7, pp. 909–996, 1988.
[35] S. Mallat, “A theory for multiresolution signal decomposition:
The wavelet representation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. PAMI-11, no. 7, pp. 674–693, July 1989.
[36] S. M. Kay, Fundamentals of Statistical Processing, Volume I: Estimation Theory. Englewood Cliffs, NJ: Prentice-Hall, 1993.
[37] M. J. Shensa, “The discrete wavelet transform: Wedding the à
trous and Mallat algorithms,” IEEE Trans. Signal Process., vol. 40,
no. 10, pp. 2464–2482, Oct. 1992.
[38] G. P. Nason and B. W. Silverman, “The stationary wavelet transform and some statistical applications,” in Wavelets and Statistics (Lecture Notes in Statistics, vol. 103), A. Antoniadis and
G. Oppenheim, Eds. New York: Springer-Verlag, 1995, pp. 281–299.
[39] F. Argenti, G. Torricelli, and L. Alparone, “Signal dependent
noise removal in the undecimated wavelet domain,” in Proc.
IEEE Int. Conf. Acoustics, Speech and Signal Processing (ICASSP),
2002, vol. 4, pp. 3293–3296.
[40] P. P. Vaidyanathan, Multirate Systems and Filter Banks, Englewood Cliffs, NJ: Prentice-Hall, 1992.
[41] M. N. Do and M. Vetterli, “The contourlet transform: An efficient directional multiresolution image representation,” IEEE
Trans. Image Process., vol. 14, no. 12, pp. 2091–2106, Dec. 2005.
[42] E. Candès and D. Donoho, “Curvelets a surprisingly effective nonadaptive representation for objects with edges,” in Curves and Surface Fitting: Saint-Malo 1999, A. Cohen, C. Rabut, and L. Schumaker,
Eds. Nashville, TN: Vanderbilt Univ. Press, 2000, pp. 105–120.
[43] J. L. Starck, E. J. Candès, and D. L. Donoho, “The curvelet transform for image denoising,” IEEE Trans. Image Process., vol. 11, no.
6, pp. 670–684, June 2002.
[44] F. Argenti, T. Bianchi, G. Martucci di Scarfizzi, and L. Alparone,
“LMMSE and MAP estimators for reduction of multiplicative
noise in the nonsubsampled contourlet domain,” Signal Process.,
vol. 89, no. 10, pp. 1891–1901, Oct. 2009.
[45] L. Alparone, F. Argenti, T. Bianchi, and A. Lapini, “An evaluation of Bayesian estimators and PDF models for despeckling
in the undecimated wavelet domain,” in Proc. SPIE SAR Image
Analysis, Modeling, and Techniques X, C. Notarnicola, Ed., 2010,
vol. 7829, pp. 782902-1–782902-12.
[46] J.-S. Lee, “Speckle analysis and smoothing of synthetic aperture
radar images,” Comput. Graph. Image Process., vol. 17, no. 1, pp.
24–32, Sept. 1981.
[47] J.-S. Lee, “Speckle suppression and analysis for synthetic aperture
radar images,” Opt. Eng., vol. 25, no. 5, pp. 636–643, May 1986.
[48] F. T. Ulaby, F. Kouyate, B. Brisco, and T. H. L. Williams, “Textural information in SAR images,” IEEE Trans. Geosci. Remote
Sensing, vol. 24, no. 2, pp. 235–245, Mar. 1986.
[49] J.-S. Lee, “Refined filtering of image noise using local statistics,” Comput. Graph. Image Process., vol. 15, no. 2, pp. 380–389,
Apr. 1981.
[50] V. S. Frost, J. A. Stiles, K. S. Shanmugan, and J. C. Holtzman, “A
model for radar images and its application to adaptive digital
filtering of multiplicative noise,” IEEE Trans. Pattern Anal. Mach.
Intell., vol. PAMI-4, no. 2, pp. 157–166, Feb. 1982.
[51] A. Lopès, R. Touzi, and E. Nezry, “Adaptive speckle filters and
scene heterogeneity,” IEEE Trans. Geosci. Remote Sensing, vol. 28,
no. 6, pp. 992–1000, Nov. 1990.
[52] D. T. Kuan, A. A. Sawchuk, T. C. Strand, and P. Chavel, “Adaptive noise smoothing filter for images with signal-dependent
noise,” IEEE Trans. Pattern Anal. Mach. Intell., vol. PAMI-7, no. 2,
pp. 165–177, Feb. 1985.
SEPTEMBER 2013
[53] D. T. Kuan, A. A. Sawchuk, T. C. Strand, and P. Chavel, “Adaptive restoration of images with speckle,” IEEE Trans. Acoust.,
Speech, Signal Process., vol. ASSP-35, no. 3, pp. 373–383, Mar. 1987.
[54] A. Lopès, E. Nezry, R. Touzi, and H. Laur, “Maximum a posteriori speckle filtering and first order texture models in SAR
images,” in Proc. IEEE Int. Geoscience and Remote Sensing Symp.
(IGARSS), 1990, pp. 2409–2412.
[55] A. Lopès, E. Nezry, R. Touzi, and H. Laur, “Structure detection
and statistical adaptive speckle filtering in SAR images,” Int. J.
Remote Sensing, vol. 14, no. 9, pp. 1735–1758, June 1993.
[56] A. Baraldi and F. Parmiggiani, “A refined gamma MAP SAR
speckle filter with improved geometrical adaptivity,” IEEE Trans.
Geosci. Remote Sensing, vol. 33, no. 6, pp. 1245–1257, Nov. 1995.
[57] M. Walessa and M. Datcu, “Model-based despeckling and information extraction from SAR images,” IEEE Trans. Geosci. Remote
Sensing, vol. 38, no. 5, pp. 2258–2269, Sept. 2000.
[58] D. E. Molina, D. Gleich, and M. Datcu, “Evaluation of Bayesian
despeckling and texture extraction methods based on Gauss–
Markov and auto-binomial Gibbs random fields: Application
to TerraSAR-X data,” IEEE Trans. Geosci. Remote Sensing, vol. 50,
no. 5, pp. 2001–2025, May 2012.
[59] B. Aiazzi, L. Alparone, and S. Baronti, “Multiresolution local-statistics speckle filtering based on a ratio Laplacian pyramid,” IEEE Trans.
Geosci. Remote Sensing, vol. 36, no. 5, pp. 1466–1476, Sept. 1998.
[60] P. Meer, R.-H. Park, and K. Cho, “Multiresolution adaptive
image smoothing,” Graph. Models Image Process., vol. 56, no. 2,
pp. 140–148, Mar. 1994.
[61] B. Aiazzi, L. Alparone, S. Baronti, and G. Borri, “Pyramid-based
multiresolution adaptive filters for additive and multiplicative image noise,” IEEE Trans. Circuits Syst. II, vol. 45, no. 8, pp.
1092–1097, Aug. 1998.
[62] B. Aiazzi, L. Alparone, S. Baronti, and F. Lotti, “Lossless image
compression by quantization feedback in a content-driven enhanced Laplacian pyramid,” IEEE Trans. Image Process., vol. 6,
no. 6, pp. 831–843, June 1997.
[63] B. Aiazzi, L. Alparone, S. Baronti, G. Chirò, F. Lotti, and M.
Moroni, “A pyramid-based error-bounded encoder: An evaluation on X-ray chest images,” Signal Process., vol. 59, no. 2, pp.
173–187, June 1997.
[64] L. Gagnon and A. Jouan, “Speckle filtering of SAR images: A
comparative study between complex-wavelet-based and standard filters,” in Proc. SPIE, Wavelet Applications in Signal and Image processing V, 1997, vol. 3169, pp. 80–91.
[65] E. Hervet, R. Fjørtoft, P. Marthon, and A. Lopès, “Comparison
of wavelet-based and statistical speckle filters,” in Proc. SPIE
SAR Image Analysis, Modelling, and Techniques III, F. Posa, Ed.,
1998, vol. 3497, pp. 43–54.
[66] M. Simard, G. DeGrandi, K. P. B. Thomson, and G. B. Bénié,
“Analysis of speckle noise contribution on wavelet decomposition of SAR images,” IEEE Trans. Geosci. Remote Sensing, vol. 36,
no. 6, pp. 1953–1962, Nov. 1998.
[67] D. L. Donoho, “Denoising by soft-thresholding,” IEEE Trans. Inform. Theory, vol. 41, no. 3, pp. 613–627, Mar. 1995.
[68] H. Guo, J. E. Odegard, M. Lang, R. A. Gopinath, I. W. Selesnick, and C. S. Burrus, “Wavelet based speckle reduction with
application to SAR based ATD/R,” in Proc. IEEE Int. Conf. Image
Processing (ICIP), 1994, vol. 1, pp. 75–79.
[69] J. R. Sveinsson and J. A. Benediktsson, “Almost translation invariant wavelet transformations for speckle reduction of SAR
images,” IEEE Trans. Geosci. Remote Sensing, vol. 41, no. 510, pp.
2404–2408, Oct. 2003.
[70] Z. Zeng and I. Cumming, “Bayesian speckle noise reduction
using the discrete wavelet transform,” in Proc. IEEE Int. Geoscience and Remote Sensing Symp. (IGARSS), 1998, vol. 1, pp. 7–9.
[71] H. Xie, L. E. Pierce, and F. T. Ulaby, “SAR speckle reduction using wavelet denoising and Markov random field modeling,”
2212, Oct. 2002.
33
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
[72] A. Achim, P. Tsakalides, and A. Bezerianos, “SAR image denoising via Bayesian wavelet shrinkage based on heavy-tailed
modeling,” IEEE Trans. Geosci. Remote Sensing, vol. 41, no. 8, pp.
1773–1784, Aug. 2003.
[73] S. Solbø and T. Eltoft, “Homomorphic wavelet-based statistical
despeckling of SAR images,” IEEE Trans. Geosci. Remote Sens.,
vol. 42, no. 4, pp. 711–721, Apr. 2004.
[74] M. I. H. Bhuiyan, M. O. Ahmad, and M. N. S. Swamy, “A new homomorphic Bayesian wavelet-based MMAE filter for despeckling SAR images,” in Proc. IEEE Int. Symp. Circuits and Systems
(ISCAS), May 2005, vol. 5, pp. 4935–4938.
[75] M. I. H. Bhuiyan, M. O. Ahmad, and M. N. S. Swamy, “Spatially
adaptive wavelet-based method using the Cauchy prior for denoising the SAR images,” IEEE Trans. Circuits Syst. Video Technol.,
vol. 17, no. 4, pp. 500–507, Apr. 2007.
[76] A. Achim, E. E. Kuruoglu, and J. Zerubia, “SAR image filtering
based on the heavy-tailed Rayleigh model,” IEEE Trans. Image
Process., vol. 15, no. 9, pp. 2686–2693, Sept. 2006.
[77] S. Foucher, G. B. Bénié, and J.-M. Boucher, “Multiscale MAP
filtering of SAR images,” IEEE Trans. Image Process., vol. 10, no.
1, pp. 49–60, Jan. 2001.
[78] F. Argenti and L. Alparone, “Speckle removal from SAR images
in the undecimated wavelet domain,” IEEE Trans. Geosci. Remote
Sensing, vol. 40, no. 11, pp. 2363–2374, Nov. 2002.
[79] M. Dai, C. Peng, A. K. Chan, and D. Loguinov, “Bayesian wavelet shrinkage with edge detection for SAR image despeckling,”
IEEE Trans. Geosci. Remote Sensing, vol. 42, no. 8, pp. 1642–1648,
Aug. 2004.
[80] R. Touzi, A. Lopès, and P. Bousquet, “A statistical and geometrical edge detector for SAR images,” IEEE Trans. Geosci. Remote
[81] S. Solbø and T. Eltoft, “C-WMAP: A statistical speckle filter operating in the wavelet domain,” Int. J. Remote Sens., vol. 25, no.
5, pp. 1019–1036, Mar. 2004.
[82] F. Argenti, T. Bianchi, and L. Alparone, “Multiresolution MAP
despeckling of SAR images based on locally adaptive generalized Gaussian PDF modeling,” IEEE Trans. Image Process., vol.
15, no. 11, pp. 3385–3399, Nov. 2006.
[83] T. Bianchi, F. Argenti, and L. Alparone, “Segmentation-based
MAP despeckling of SAR images in the undecimated wavelet
domain,” IEEE Trans. Geosci. Remote Sensing, vol. 46, no. 9, pp.
2728–2742, Sept. 2008.
[84] R. Tao, H. Wan, and Y. Wang, “Artifact-free despeckling of SAR
images using contourlet,” IEEE Geosci. Remote Sensing Lett., vol.
9, no. 5, pp. 980–984, Sept. 2012.
[85] F. Argenti, T. Bianchi, A. Lapini, and L. Alparone, “Fast MAP
despeckling based on Laplacian–Gaussian modeling of wavelet
coefficients,” IEEE Geosci. Remote Sensing Lett., vol. 9, no. 1, pp.
13–17, Jan. 2012.
[86] H. Chen, Y. Zhang, H. Wang, and C. Ding, “Stationary-waveletbased despeckling of SAR images using two-sided generalized
gamma models,” IEEE Geosci. Remote Sensing Lett., vol. 9, no. 6,
pp. 1061–1065, Nov. 2012.
[87] H.-C. Li, W. Hong, Y.-R. Wu, and P.-Z. Fan, “Bayesian wavelet
shrinkage with heterogeneity-adaptive threshold for SAR image despeckling based on generalized gamma distribution,”
2402, Apr. 2013.
[88] A. Pižurica, W. Philips, I. Lemahieu, and M. Acheroy, “A versatile
wavelet domain noise filtration technique for medical imaging,”
IEEE Trans. Med. Imag., vol. 22, no. 3, pp. 323–331, Mar. 2003.
[89] S. Fukuda and H. Hirosawa, “Smoothing effect of waveletbased speckle filtering: The Haar basis case,” IEEE Trans. Geosci. Remote Sensing, vol. 37, no. 2, pp. 1168–1172, Mar. 1999.
[90] Y. Hawwar and A. Reza, “Spatially adaptive multiplicative noise
image denoising technique,” IEEE Trans. Image Process., vol. 11,
no. 12, pp. 1397–1404, Dec. 2002.
34
[91] L. Alparone, S. Baronti, and R. Carlà, “Two-dimensional rankconditioned median filter,” IEEE Trans. Circuits Syst. II, vol. 42,
no. 2, pp. 130–132, Feb. 1995.
[92] L. Alparone, S. Baronti, R. Carlà, and C. Puglisi, “An adaptive
order-statistics filter for SAR images,” Int. J. Remote Sens., vol. 17,
no. 7, pp. 1357–1365, May 1996.
[93] T. R. Crimmins, “Geometric filter for speckle reduction,” Appl.
Opt., vol. 24, no. 10, pp. 1438–1443, May 1985.
[94] L. Alparone and A. Garzelli, “Decimated geometric filter for
edge-preserving removal of non-white image noise,” Pattern
Recogn. Lett., vol. 19, no. 1, pp. 89–96, Jan. 1998.
[95] P. Perona and J. Malik, “Scale-space and edge detection using
anisotropic diffusion,” IEEE Trans. Pattern Anal. Mach. Intell.,
vol. 12, no. 7, pp. 629–639, July 1990.
[96] Y. Yu and S. T. Acton, “Speckle reducing anisotropic diffusion,” IEEE Trans. Image Process., vol. 11, no. 11, pp. 1260–1270,
Nov. 2002.
[97] Y. Yu and S. T. Acton, “Automated delineation of coastline from
polarimetric SAR imagery,” Int. J. Remote Sens., vol. 25, no. 17,
pp. 3423–3438, Sept. 2004.
[98] R. G. White, “A simulated annealing algorithm for SAR and MTI
image cross section estimation,” in Proc. SPIE SAR Data Processing for Remote Sensing 137, 1994, vol. 2316, pp. 339–360.
[99] I. McConnell and C. Oliver, “Comparison of annealing and iterated filters for speckle reduction in SAR,” in Proc. SPIE Microwave
Sensing and Synthetic Aperture Radar 74, 1994, vol. 2958, pp. 74–85.
[100] J. Schou and H. Skriver, “Restoration of polarimetric SAR images using simulated annealing,” IEEE Trans. Geosci. Remote
Sensing, vol. 39, no. 9, pp. 2005–2016, Sept. 2001.
[101] J.-S. Lee, “Digital image smoothing and the sigma filter,” Comput.
Vis. Graph. Image Process., vol. 24, no. 2, pp. 255–269, Nov. 1983.
[102] J.-S. Lee, “A simple speckle smoothing algorithm for synthetic
aperture radar images,” IEEE Trans. Syst., Man, Cybern., vol.
SMC-13, no. 1, pp. 85–89, Jan. 1983.
[103] L. Alparone, S. Baronti, and A. Garzelli, “A hybrid sigma filter
for unbiased and edge-preserving speckle reduction,” in Proc.
IEEE Int. Geoscience and Remote Sensing Symp. (IGARSS), 1995,
vol. 2, pp. 1409–1411.
[104] J.-S. Lee, J.-H. Wen, T. L. Ainsworth, K.-S. Chen, and A. J.
Chen, “Improved sigma filter for speckle filtering of SAR imagery,” IEEE Trans. Geosci. Remote Sensing, vol. 47, no. 1, pp.
202–213, Jan. 2009.
[105] C. Tomasi and R. Manduchi, “Bilateral filtering for gray and
color images,” in Proc. 6th Int. Conf. Computer Vision (ICCV),
1998, pp. 839–846.
[106] W. G. Zhang, Q. Zhang, and C. S. Yang, “Improved bilateral
filtering for SAR image despeckling,” Electron. Lett., vol. 47, no.
4, pp. 286–288, Feb. 2011.
[107] G.-T. Li, C.-L. Wang, P.-P. Huang, and W.-D. Yu, “SAR image
despeckling using a space-domain filter with alterable window,” IEEE Geosci. Remote Sensing Lett., vol. 10, no. 2, pp. 263–
267, Mar. 2013.
[108] C.-A. Deledalle, L. Denis, and F. Tupin, “Iterative weighted
maximum likelihood denoising with probabilistic patchbased weights,” IEEE Trans. Image Process., vol. 18, no. 12, pp.
2661–2672, Dec. 2009.
[109] S. Parrilli, M. Poderico, C. V. Angelino, and L. Verdoliva, “A
nonlocal SAR image denoising algorithm based on LLMMSE
wavelet shrinkage,” IEEE Trans. Geosci. Remote Sensing, vol. 50,
no. 2, pp. 606–616, Feb. 2012.
[110] A. Buades, B. Coll, and J.-M. Morel, “A non-local algorithm for
image denoising,” in Proc. IEEE Conf. Computer Vision and Pattern Recognition (ICCVPR), 2005, vol. 2, pp. 60–65.
[111] K. Dabov, A. Foi, V. Katkovnik, and K. Egiazarian, “Image
denoising by sparse 3-D transform-domain collaborative
filtering,” IEEE Trans. Image Process., vol. 16, no. 8, pp. 2080–
2095, Aug. 2007.
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
[112] T. Teuber and A. Lang, “A new similarity measure for nonlocal
filtering in the presence of multiplicative noise,” Comput. Stat.
Data Anal., vol. 56, no. 12, pp. 3821–3842, Dec. 2012.
[113] C. Kervrann, J. Boulanger, and P. Coupé, “Bayesian nonlocal means filter, image redundancy and adaptive dictionaries for noise removal,” in Proc. 1st Int. Conf. on Scale Space
and Variational Methods in Computer Vision (SSVM), 2007, pp.
520–532.
[114] P. Coupe, P. Hellier, C. Kervrann, and C. Barillot, “Bayesian
non local means-based speckle filtering,” in Proc. 5th IEEE
Int. Symp. Biomedical Imaging: From Nano to Macro, 2008, pp.
1291–1294.
[115] H. Zhong, Y. Li, and L. Jiao, “SAR image despeckling using
Bayesian non-local means filter with sigma preselection,” IEEE
Geosci. Remote Sensing Lett., vol. 8, no. 4, pp. 809–813, July 2011.
[116] D. Gragnaniello, G. Poggi, and L. Verdoliva, “Classificationbased nonlocal SAR despeckling,” in Proc. Tyrrhenian Workshop
on Advances in Radar and Remote Sensing, 2012, pp. 121–125.
[117] L. I. Rudin, S. Osher, and E. Fatemi, “Nonlinear total variation
based noise removal algorithms,” Physica D, vol. 60, nos. 1–4,
pp. 259–268, Nov. 1992.
[118] L. I. Rudin, P.-L. Lions, and S. Osher, “Multiplicative denoising
and deblurring: Theory and algorithms,” in Geometric Level Set
Methods in Imaging, Vision, and Graphics. New York: SpringerVerlag, 2003, pp. 103–119.
[119] G. Aubert and J. Aujol, “A variational approach to removing
multiplicative noise,” SIAM J. Appl. Math., vol. 68, no. 4, pp.
925–946, Dec. 2008.
[120] J. Shi and S. Osher, “A nonlinear inverse scale space method for a
convex multiplicative noise model,” SIAM J. Imaging Sci., vol. 1,
no. 3, pp. 294–321, Sept. 2008.
[121] Y.-M. Huang, M. K. Ng, and Y.-W. Wen, “A new total variation
method for multiplicative noise removal,” SIAM J. Imaging Sci.,
vol. 2, no. 1, pp. 20–40, Jan. 2009.
[122] S. Durand, J. Fadili, and M. Nikolova, “Multiplicative noise removal using L1 fidelity on frame coefficients,” J. Math. Imaging
Vis., vol. 36, no. 3, pp. 201–226, Mar. 2010.
[123] G. Steidl and T. Teuber, “Removing multiplicative noise by
Douglas-Rachford splitting methods,” J. Math. Imaging Vis.,
vol. 36, no. 2, pp. 168–184, Feb. 2010.
[124] J. M. Bioucas-Dias and M. A. T. Figueiredo, “Multiplicative
noise removal using variable splitting and constrained optimization,” IEEE Trans. Image Process., vol. 19, no. 7, pp. 1720–
1730, July 2010.
[125] L. Denis, F. Tupin, J. Darbon, and M. Sigelle, “SAR image regularization with fast approximate discrete minimization,”
IEEE Trans. Image Process., vol. 18, no. 7, pp. 1588–1600,
July 2009.
[126] F. Palsson, J. R. Sveinsson, M. O. Ulfarsson, and J. A. Benediktsson, “SAR image denoising using total variation based
regularization with SURE-based optimization of the regularization parameter,” in Proc. IEEE Int. Geoscience and Remote
Sensing Symp. (IGARSS), 2012, pp. 2160–2163.
[127] D. Gleich and M. Kseneman, “A comparison of regularizationbased methods for despeckling of SLC SAR images,” in Proc.
9th European Conf. Synthetic Aperture Radar (EUSAR), 2012, pp.
784–787.
[128] D. L. Donoho, “Compressed sensing,” IEEE Trans. Inform. Theory, vol. 52, no. 4, pp. 1289–1306, Apr. 2006.
[129] S. Foucher, “SAR image filtering via learned dictionaries and
sparse representations,” in Proc. IEEE Int. Geoscience and Remote
Sensing Symp. (IGARSS), 2008, vol. I, pp. 229–232.
SEPTEMBER 2013
[130] M. Yang and G. Zhang, “SAR image despeckling using overcomplete dictionary,” Electron. Lett., vol. 48, no. 10, pp. 596–
597, May 2012.
[131] Y. Hao, X. Feng, and J. Xu, “Multiplicative noise removal via
sparse and redundant representations over learned dictionaries and total variation,” Signal Process., vol. 92, no. 6, pp. 1536–
1549, June 2012.
[132] R. Coifman and D. L. Donoho, “Translation-invariant de-noising,” in Wavelets and Statistics (Lecture Notes), A. Antoniadis,
Ed. Berlin, Heidelberg: Springer-Verlag, 1995.
[133] F. Argenti, T. Bianchi, A. Lapini, and L. Alparone, “Bayesian
despeckling of SAR images based on Laplacian–Gaussian
modeling of undecimated wavelet coefficients,” in Proc. IEEE
Int. Conf. Acoustics, Speech and Signal Processing (ICASSP), 2011,
pp. 1445–1448.
[134] H. Rabbani, M. Vafadust, P. Abolmaesumi, and S. Gazor,
“Speckle noise reduction of medical ultrasound images in
complex wavelet domain using mixture priors,” IEEE Trans.
Biomed. Eng., vol. 55, no. 9, pp. 2152–2160, Sept. 2008.
[135] Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli,
“Image quality assessment: From error visibility to structural similarity,” IEEE Trans. Image Process., vol. 13, no. 4, pp.
600–612, Apr. 2004.
[136] F. Sattar, L. Floreby, G. Salomonsson, and B. Lovstrom,
“Image enhancement based on a nonlinear multiscale
method,” IEEE Trans. Image Process., vol. 6, no. 6, pp. 888–
895, June 1997.
[137] T. Bianchi, F. Argenti, A. Lapini, and L. Alparone, “Amplitude
vs intensity despeckling in the wavelet domain using Bayesian
estimators,” in Proc. Tyrrhenian Workshop on Advances in Radar
and Remote Sensing, 2012, pp. 267–274.
[138] G. R. Benitz, “High-definition vector imaging,” Lincoln Lab. J.,
vol. 10, no. 2, pp. 147–170, 1997.
[139] M. Çetin, W. C. Karl, and D. A. Castañon, “Evaluation of a
regularized SAR imaging technique based on recognitionoriented features,” in Proc. SPIE Algorithms for Synthetic Aperture
Radar Imagery VII, 2000, vol. 4053, pp. 40–51.
[140] A. Lapini, T. Bianchi, F. Argenti, and L. Alparone, “Blind speckle decorrelation for SAR image despeckling,” IEEE Trans. Geosci.
Remote Sensing, to be published.
[141] A. Lapini, T. Bianchi, F. Argenti, and L. Alparone, “Blind whitening of correlated speckle to enforce despeckling of singlelook high-resolution SAR images,” in Proc. SPIE Image and
Signal Processing for Remote Sensing XVIII, 2012, vol. 8537, pp.
85370Z–85370Z-8.
[142] B. Aiazzi, L. Alparone, F. Argenti, S. Baronti, T. Bianchi, and A.
Lapini, “An experimental setup for multiresolution despeckling of COSMO-SkyMed image products,” in Proc. SPIE SAR
Image Analysis, Modeling, and Techniques XII, 2011, vol. 8536,
pp. 853603–853603-6.
[143] T. Bianchi, F. Argenti, A. Lapini, and L. Alparone, “Amplitude
vs intensity Bayesian despeckling in the wavelet domain for
SAR images,” Digit. Signal Process., vol. 23, no. 5, pp. 1353–
1362, Sept. 2013.
[144] B. Aiazzi, L. Alparone, F. Argenti, S. Baronti, T. Bianchi, and
A. Lapini, “An unsupervised method for quality assessment
of despeckling: An evaluation on COSMO-SkyMed data,” in
Proc. SPIE, 2011, vol. 8197, pp. 81790D–81790D-10.
[145] B. Aiazzi, L. Alparone, S. Baronti, A. Garzelli, and C. Zoppetti,
“Nonparametric change detection in multitemporal SAR images based on mean-shift clustering,” IEEE Trans. Geosci. Remote Sensing, vol. 51, no. 4, pp. 2022–2031, Apr. 2013. GRS
35
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
TECHNICAL COMMITTEES
FABIO PACIFICI, DigitalGlobe, Inc., USA
QIAN DU, Mississippi State University, USA
SAURABH PRASAD, University of Houston, USA
Report on the 2013 IEEE GRSS Data Fusion
Contest: Fusion of Hyperspectral and LiDAR Data
T
he Data Fusion Contest is organized by the Data
Fusion Technical Committee of the Geoscience
and Remote Sensing Society (GRSS). The Committee
serves as a global, multi-disciplinary, network for geospatial data fusion, with the
aim of connecting people and resources,
educating students and professionals, and
promoting the best practices in data fusion
applications.
The Contest has been held annually
since 2006. It is open not only to IEEE
members, but to everyone, with the goal of evaluating
existing methodologies at the research or operational
level to solve remote sensing problems using data from
a variety of sensors.
1. OVERVIEW OF PREVIOUS DATA FUSION CONTESTS
The focus of the 2006 Contest was on the fusion of multispectral and panchromatic images [1]. Six simulated
Pleiades images were provided by the French National
Space Agency (CNES). Each data set included a very
high spatial resolution panchromatic image (0.80 m
resolution) and its corresponding multispectral image
(3.2 m resolution). A high spatial resolution multispectral image was available as ground reference, which
was used by the organizing committee for evaluation
but not distributed to the participants.
In 2007, the Contest theme was urban mapping using
synthetic aperture radar (SAR) and optical data, and
9 ERS amplitude data sets and 2 Landsat multispectral
images were made available [2]. The task was to obtain
a classification map as accurate as possible with respect
to the unknown (to the participants) ground reference,
depicting land cover and land use patterns for the urban
area under study.
36
The 2008 Contest was dedicated to the classification of
very high spatial resolution (1.3 m) hyperspectral imagery
[3]. The task was again to obtain a classification map as
accurate as possible with respect to the unknown
(to the participants) ground reference. The data
set was collected by the Reflective Optics System
Imaging Spectrometer (ROSIS-03) optical sensor
with 115 bands covering the 0.43–0.86 μm spectral range.
In 2009–2010, the aim of Contest was to perform change detection using multi-temporal and
multi-modal data [4]. Two pairs of data sets were available over Gloucester, UK, before and after a flood event.
The data set contained SPOT and ERS images (before and
after the disaster). The optical and SAR images were provided by CNES. Similar to previous years’ Contests, the
ground truth used to assess the results was not provided
to the participants. Each set of results was tested and
ranked a first time using the Kappa coefficient. The best
five results were used to perform decision fusion with
majority voting. Then, re-ranking was carried out after
evaluating the level of improvement with respect to the
fusion results.
A set of WorldView-2 multi-angular images was provided by DigitalGlobe for the 2011 Contest [5]. This
unique set was composed of five Ortho Ready Standard
multi-angular acquisitions, including both 16 bit panchromatic and multispectral 8-band images. The data
were collected over Rio de Janeiro (Brazil) in January
2010 within a three minute time frame with satellite
elevation angles of 44.7°, 56.0°, and 81.4° in the forward direction, and 59.8° and 44.6° in the backward
direction. Since there were a large variety of possible
applications, each participant was allowed to decide a
research topic to work on, exploring the most creative
use of optical multi-angular information. At the end of
the Contest, each participant was required to submit a
paper describing in detail the problem addressed, the
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
FIGURE 1. Composition of the hyperspectral and LiDAR data
sets over the University of Houston campus.
method used, and the final result generated. The papers
submitted were automatically formatted to hide names and
affiliations of the authors to ensure neutrality and impartiality of the reviewing process.
The 2012 Contest was designed to investigate the potential of multi-modal/multi-temporal fusion of very high spatial
resolution imagery in various remote sensing applications [6].
Three different types of data sets (optical, SAR, and LiDAR) over
downtown San Francisco were made available by DigitalGlobe,
Astrium Services, and the United States Geological Survey
(USGS), including QuickBird, WorldView-2, TerraSAR-X, and
LiDAR imagery. The image scenes covered a number of large
buildings, skyscrapers, commercial and industrial structures,
a mixture of community parks and private housing, and highways and bridges. Following the success of the multi-angular
Data Fusion Contest in 2011, each participant was again
required to submit a paper describing in detail the problem
addressed, method used, and final results generated for review.
buildings, highways, railway, and vehicles. The validation
samples that the Contest organizers used to evaluate the submissions were not disclosed.
Best Paper Challenge, with the objective of promoting novel use of hyperspectral and LiDAR data. The deliverable was a 4-page manuscript describing the problem,
methodology, results and discussion. The goal of this challenge was to encourage the participants to consider hyperspectral and LiDAR data fusion problems and to demonstrate novel and effective approaches to address them.
The Data Fusion Award Committee consisted of seven
independent judges from universities and industries:
◗ Jocelyn Chanussot, Grenoble Institute of Technology,
France
◗ Melba Crawford, Purdue University, USA
◗ Jenny (Qian) Du, Mississippi State University, USA
◗ Paolo Gamba, University of Pavia, Italy
◗ Fabio Pacifici, DigitalGlobe, Inc., USA
◗ Antonio Plaza, University of Extremadura, Spain
◗ Saurabh Prasad, University of Houston, USA
Papers were judged in terms of sound scientific reasoning, problem definition, methodology, validation, and
presentation.
3. OUTCOME OF THE CONTEST
More than 900 researchers from universities, national
labs, space agencies, and corporations across the globe registered to the Contest, demonstrating the great interest of
the community in the DFTC activities in promoting cutting-edge research of remote sensing image processing and
analysis. The data sets were downloaded from a total of
69 different countries, with a large number of registrations
from less developed areas. Fig. 2 shows the geographical
distribution of the participants, where other indicates the
sum of all countries with less than 10 participants.
2. 2013 DATA FUSION CONTEST
The 2013 Contest was aimed at exploring the synergetic use
of hyperspectral and LiDAR data. The hyperspectral imagery was composed of 144 spectral bands from 380 to 1050
nm. A co-registered LiDAR derived Digital Surface Model
(DSM) was also made available to all participants. Both data
sets had the same spatial resolution (2.5 m). As shown in
Fig. 1, the data was acquired by the National Science Foundation (NSF)-funded Center for Airborne
Laser Mapping (NCALM) in the summer
of 2012 over the University of Houston
USA
17%
and the neighboring urban area. The data
17%
China
pre-processing was conducted by student
India
volunteers at UH’s Hyperspectral Image
Iran
1%
Canada
Analysis group, and NCALM staff. A
1%
1%
France
ground truth corresponding to this data1%
Germany
2%
set was created by the contest organizing
Italy
17%
2%
committee via photo-interpretation.
Pakistan
This year, the Contest consisted of
2%
Spain
two parallel competitions:
Turkey
2%
Best Classification Challenge, with the
Japan
Belgium
objective of promoting innovation in clas2%
Malaysia
sification algorithms, and to provide fair
3%
Egypt
performance comparisons among state3%
Greece
14%
of-the-art algorithms. For this task, users
Other
4%
were provided with training samples from
8%
14 classes of interest, including various
types of vegetation, soil, water, but also FIGURE 2. Geographical distribution of the participants for countries with more than
less common targets, such as commercial 10 participants.
SEPTEMBER 2013
37
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
(a)
(b)
FIGURE 3. The Data Fusion Technical Committee congratulates the winners of the 2013 Contest during the Chapters and Technical Committees Dinner at IGARSS 2013. From left to right in each photo: (a) Qian Du, Fabio Pacifici, Andreas Merentitis, and Saurabh Prasad; (b) Qian Du,
Fabio Pacifici, Paul Scheunders (on behalf of Liao et al.), and Saurabh Prasad.
You can contact the Committee Chairs by e-mail at: _______
[email protected].
If you are interested in joining the Data Fusion Technical Committee, please
send an e-mail with:
◗ First and Last Name
◗ Institution/Company
◗ Country
◗ IEEE Membership Number (if available)
◗ E-mail
Members receive information regarding Data Fusion research and applications, and updates on the annual Data Fusion Contest. Membership in the Data
Fusion Technical Committee is free!
Join the LinkedIn IEEE GRSS Data Fusion Discussion Forum: http://www.
linkedin.com/groups/IEEE-GRSS-Data-Fusion-Discussion-3678437.
comes will be submitted for peer review to the IEEE Journal of
Selected Topics in Applied Earth Observations and Remote Sensing
(JSTARS). To further enhance its impact in the community,
the Data Fusion Technical Committee will support its openaccess publication cost. Funding provided by the IEEE GRSS
and DigitalGlobe, Inc.
ACKNOWLEDGMENTS
The IEEE GRSS Data Fusion Technical Committee would like
to express its great appreciation to NCALM for providing the
data, to the Hyperspectral Image Analysis group, NCALM
staff, both at UH, for preparing the data, and to GRSS and
DigitalGlobe Inc. for their continuous support in providing
funding and resources for the Data Fusion Contest.
REFERENCES
Final results were announced at the 2013 IEEE International Geoscience and Remote Sensing Symposium held
in Melbourne, Australia. The winners of the 2013 Data
Fusion Contest are:
◗ Christian Debes1, Andreas Merentitis1, Roel Heremans1, Jürgen Hahn2, Nick Frangiadakis1, and Tim van
Kasteren1 from AGT International and Technische Universität Darmstadt, Germany, are the winners of the Best
Classification Challenge. Details can be found at http://
____
hyperspectral.ee.uh.edu/?page_id=695.
_________________________
◗ Wenzhi Liao, Rik Bellens, Aleksandra Pizurica, Sidharta
Gautama, and Wilfried Philips from Ghent University,
Belgium, are the winners of the Best Paper Challenge with
a paper entitled “Graph-Based Feature Fusion of Hyperspectral and Lidar Remote Sensing Data Using Morphological Features”. Details can be found at http://hyper________
spectral.ee.uh.edu/?page_id=795.
_____________________
As shown in Fig. 3, the winning teams were awarded IEEE
GRSS Certificates of Appreciations and iPads during the
Technical Committees and Chapters Dinner. Additionally,
as is tradition, a manuscript summarizing the Contest out38
[1] L. Alparone, L. Wald, J. Chanussot, C. Thomas, P. Gamba, and L.
M. Bruce, “Comparison of pansharpening algorithms: Outcome
of the 2006 GRS-S Data Fusion Contest,” IEEE Trans. Geosci.
Remote Sensing, vol. 45, no. 10, pp. 3012–3021, Oct. 2007.
[2] F. Pacifici, F. Del Frate, W. J. Emery, P. Gamba, and J. Chanussot,
“Urban mapping using coarse SAR and optical data: Outcome of
the 2007 GRS-S Data Fusion Contest,” IEEE Geosci. Remote Sensing Lett., vol. 5, no. 3, pp. 331–335, July 2008.
[3] G. Licciardi, F. Pacifici, D. Tuia, S. Prasad, T. West, F. Giacco, J.
Inglada, E. Christophe, J. Chanussot, and P. Gamba, “Decision
fusion for the classification of hyperspectral data: Outcome of
the 2008 GRS-S Data Fusion Contest,” IEEE Trans. Geosci. Remote
[4] N. Longbotham, F. Pacifici, T. Glenn, A. Zare, M. Volpi, D. Tuia, E. Christophe, J. Michel, J. Inglada, J. Chanussot, and Q. Du, “Multi-modal
change detection, application to the detection of flooded areas: Outcome of the 2009-2010 Data Fusion Contest,” IEEE J. Select. Top. Appl.
Earth Observ. Remote Sensing, vol. 5, no. 1, pp. 331–342, Feb. 2012.
[5] F. Pacifici and Q. Du, “Foreword to the special issue on optical
multiangular data exploitation and outcome of the 2011 GRSS
Data Fusion Contest,” IEEE J. Select. Top. Appl. Earth Observ.
Remote Sensing, vol. 5, no. 1, pp. 3–7, Feb. 2012.
[6] C. Berger, M. Voltersen, R. Eckardt, J. Eberle, T. Heyer, N. Salepci,
S. Hese, C. Schmullius, J. Tao, S. Auer, R. Bamler, K. Ewald, M.
Gartley, J. Jacobson, A. Buswell, Q. Du, and F. Pacifici. (2013,
June). Multi-modal and multi-temporal data fusion: Outcome of
the 2012 GRSS Data Fusion Contest. IEEE J. Select. Top. Appl. Earth
Observ. Remote Sensing. [Online]. 6(3), pp. 1324–1340. Available:
http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=
6480856&isnumber=6541987
GRS
_________________
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
CHAPTERS
KADIM TAŞDEMİR
IEEE GRSS Turkey Chapter
R
emote sensing has been a key element for environmental sustainability, protection and security
of the citizens, disaster management and urban monitoring. It is an indispensable tool particularly for Turkey which is one of the major agriculture states in the
region with a population of nearly seventy million
(more than 70% live in the urban cities) and which
often significantly suffers from disasters such as earthquakes, floods, landslides, and forest fires. Therefore,
remote sensing technologies have been selected as one
of the technological areas with high priority for the
Vision 2023 (the state goals for the centennial celebration of the Republic of Turkey). Specific state departments and research centers have been established to
promote and advance space technologies at every scale
from sensor development to launcher systems, including image information mining. Based on these initiatives, a national satellite with high spatial resolution,
Göktürk-2, was launched early this year, and new satellite missions are in progress.
Thanks to the increasing interest in geoscience and
remote sensing, IEEE GRSS Turkey chapter was established this year with two main motivating factors: to
serve the GRSS members, researchers and general public
interested in geoscience and remote sensing in Turkey,
through scientific, technical and educational activities; and to promote a high level of excellence among
the GRSS members in Turkey by information exchange
through meetings, workshops, and conferences. In addition, this chapter will be a good platform for building
partnerships to address the challenges in the remotesensing research. Considering the vital priorities of
Turkey’s special conditions such as the evaluation of
agricultural and natural resources, disaster preparedness, and mineral exploration possibilities, this chapter
SEPTEMBER 2013
Professor Lorenzo Bruzzone and executive committee of the
GRSS Turkey Chapter (left to right: E. Erten, K. Tasdemir,
L. Bruzzone, and G. Taskin Kaya).
Professor Lorenzo Bruzzone, during his presentation at ITU.
39
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Professor Melba Crawford and Kadim Tasdemir at RAST’2013, Istanbul.
is also of great importance for advancing geoscience and
remote sensing applications in a collaborative manner.
As a kick-off to start our chapter activities, we hosted
Professor Lorenzo Bruzzone through GRSS Distinguished
Lecturer Program at Istanbul Technical University at the
end of April 2013. Professor Bruzzone gave a talk entitled
“Current Scenario and Challenges in the Analysis of Multitemporal Remote Sensing Images” to an audience of
nearly a hundred people composed of students, faculty and
researchers. He also shared his valuable ideas on possible
GRSS activities and gave productive suggestions on remotesensing image analysis. This program also set the scene to
form collaborative agreements which have initiated projects on disaster management. In April, a special session,
cochaired by Begum Demir from University of Trento and
Kadim Taşdemir from Antalya International University, on
“Remote Sensing Image Analysis” was also organized in
Signal Processing and Communications Applications Conference (Sinyal ∙l şleme ve ∙lletişim Uygulamaları Kurultayı,
Sl∙U 2013) and fourteen studies on different aspects of
remote sensing were presented.
The sixth of the biannual international conference on
recent advancements on space technologies (RAST’2013),
technically cosponsored by GRSS, was held in Istanbul.
Professor Melba Crawford was an invited speaker and
informed the attendees about “IEEE GRSS and Remote
Sensing Technologies: Accomplishments and Challenges
for the next Decade” and gave a presentation on hyperspectral image processing. After 10 years of its first organization
in 2003, the conference sponsored by main remote sensing societies has become internationally acknowledged
on space technologies and received paper from more than
thirty countries. The next one will also be organized in
Istanbul, June 2015. Detailed information on RAST series
can be found at http://www.rast.org.tr.
The initiative of forming the chapter, started during the
IGARSS 2012 in Munich, has already been paying off by
doubling the number of GRSS members in Turkey in less
than a year. The elected executive committee members of
the GRSS Turkey Chapter are: Chair Kadim Taşdemir (Antalya International University), Vice-Chair Esra Erten (Istanbul Technical University), Secretary Gül∙şen Taşkın Kaya
(Istanbul Technical University), and Treasurer Mehmet
Kurum (TUBITAK). For more information on the chapter
activities, please contact: ___________
[email protected].
GRS
GRSS CHAPTERS AND CONTACT INFORMATION
CHAPTER LOCATION
JOINT WITH (SOCIETIES)
CHAPTER CHAIR
E-MAIL ADDRESS
Boston Section, MA
GRS
William Blackwell
[email protected]
________
Springfield Section, MA
AP, MTT, ED, GRS, LEO
Paul Siqueira
_____________
[email protected]
Western New York
GRS
Jan van Aardt
___________
[email protected]
GRS
Miguel Roman
[email protected]
_______________
Region 1: Northeastern USA
Region 2: Eastern USA
Washington, DC & Northern VA area
Region 3: Southeastern USA
Atlanta Section, GA
AES, GRS
Clayton Kerce
________________
[email protected]
Eastern North Carolina Section
GRS
Linda Hayden
[email protected]
______________
Region 4: Central USA
Central Illinois Section
LEO, GRS
Weng Cho Chew
[email protected]
___________
Southeastern Michigan Section
GRS
Adib Y. Nashashibi
[email protected]
____________
Denver Section, CO
AP, MTT, GRS
Michael Janezic
[email protected]
_____________
Houston Section, TX
AP, MTT, GRS, LEO
Christi Madsen
[email protected]
____________
Region 5: Southwestern USA
40
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
CHAPTER LOCATION
JOINT WITH (SOCIETIES)
CHAPTER CHAIR
E-MAIL ADDRESS
Region 6: Western USA
Alaska Section, AK
GRS
Franz Meyer
[email protected]
___________
Los Angeles Section, CA
GRS
Paul A. Rosen
______________
[email protected]
Region 7: Canada
Ottawa Section, ON
OE, GRS
Yifeng Zhou
[email protected]
____________
Quebec Section, Quebec, QC
AES, OE, GRS
Xavier Maldague
[email protected]
____________
Toronto Section, ON
SP, VT, AES, UFF, OE, GRS
Sri Krishnan
[email protected]
_____________
Vancouver Section, BC
AES, GRS
David G. Michelson
[email protected]
____________
Steven McClain
[email protected]
_____________
Region 8: Europe, Middle East and Africa
Benelux Section
AES, GRS
Mark Bentum
[email protected]
_____________
Croatia Section
AES, GRS
Juraj Bartolic
[email protected]
___________
France Section
GRS
Mathieu Fauvel
[email protected]
_____________
Germany Section
GRS
Irena Hajnsek
[email protected]
___________
Islamabad Section, Pakistan
GRS, AES
M. Umar Khattak
_____________
[email protected]
Italy Section
GRS
Simonetta Paloscia
[email protected]
___________
Russia Section
GRS
Anatolij Shutko
[email protected]
_________________
Saudi Arabia Section
GRS
Yakoub Bazi
[email protected]
__________
South Africa Section
AES, GRS
Meena Lysko
[email protected]
__________
[email protected]
_________
South Italy Section
GRS
Maurizio Migliaccio
[email protected]
________________
Spain Section
GRS
Antonio J. Plaza
[email protected]
_________
Student Branch, Spain Section
GRS
Pablo Benedicto
[email protected]
_____________
Turkey Section
GRS
Kadim Tasdemir
__________
[email protected]
Ukraine Section
AP, MTT, ED, AES, GRS, NPS
Kostyantyn V. Ilyenko
[email protected]
____________
United Kingdom & Rep. of Ireland (UKRI) Section
GRS, OE
Yong Xue
[email protected]
_____________
Region 9: Latin America
Student Branch, Colombia Section
GRS
Leyini Parra Espitia
___________
[email protected]
Student Branch, South Brazil Section
GRS
Marcus Vasconcelos
[email protected]
___________
Guadalajara Section, Mexico
GRS
Iván Villalón
[email protected]
____________
Australian Capital Territory and New South
Wales Sections, Australia
GRS
Xiuping Jia
[email protected]
_________
Bangalore Section, India
GRS
Daya Sagar Behara
[email protected]
____________
Beijing Section, China
GRS
Ji Wu
[email protected]
_________
Delhi Section, India
GRS
O.P.N. Calla
[email protected]
___________
Gujarat Section, India
GRS
Shiv Mohan
_______________
[email protected]
Region 10: Asia and Pacific
Indonesia Section
GRS, AES
Arifin Nugroho
[email protected]
______________
Japan Section
GRS
Yoshihisa Hara
Hara.Yoshihisa@
_________
cb.MitsubishiElectric.co.jp
______________
Malaysia Section
GRS, AES
Voon-Chet Koo
[email protected]
___________
Melbourne Section
GRS, AES
William Junek
[email protected]
_____________
Nanjing Section, China
GRS
Feng Jiao
[email protected]
_______________
Seoul Section, Korea
GRS
Joong-Sun Won
[email protected]
___________
Singapore Section
AES, GRS
See Kye Yak
[email protected]
__________
Taipei Section, Taiwan
GRS
Yang-Lang Chang
[email protected]
___________
Abbreviation Guide for IEEE Technical Societies
AES
AP
ED
EMB
LEO
MTT
SEPTEMBER 2013
Aerospace and Electronic Systems Society
Antennas and Propagation Society
Electron Devices Society
Engineering in Medicine and Biology
Lasers & Electro-Optics Society
Microwave Theory and Techniques Society
NPS
OE
SP
UFF
VT
Nuclear and Plasma Sciences Society
Oceanic Engineering Society
Signal Processing Society
Ultrasonics, Ferroelectrics, and Frequency Control Society
Vehicular Technology Society
41
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
EDUCATION
MICHAEL INGGS, University of Cape Town, South Africa
XIUPING JIA, University of New South Wales
SIMON JONES, Royal Melbourne Institute of Technology
KIM LOWELL, University of Melbourne
The GRSS Summer School Melbourne 2013
I. SUMMER SCHOOL OVERVIEW
he second deployment of the GRSS Summer School
(GR4S) became a Winter School in Melbourne, Australia, held successfully on 18–19 July at a beautiful lecture theatre of RMIT University, Melbourne, Australia.
The event was opened by Professor Simon Jones, RMIT,
Australia, and chaired by Dr. Xiuping Jia, University of
New South Wales, Australia. There were 50 attendees
(including speakers).
T
II. GOALS OF GR4S
The GRSS Summer School, first held before the Munich
IGARSS (2012) was designed as a pre-conference event
that provides an outstanding opportunity to obtain
a broad knowledge of the fundamentals of a range of
image data and analysis, from leading Australian and
international authorities, and to meet other attendees
before the conference begins.
The level of the GR4S is introductory, aimed at
early Ph.D. students, or, advanced Masters students.
This is to be contrasted with the Tutorials that are held
the Sunday for an IGARSS starts.
These Tutorials are intended to
provide advanced updates to the
Geoscience and Remote Sensing
practitioner i.e. Continuing Professional Development.
A further objective of the GR4S
is to allow students to meet some
of the well known local and international experts in the field of
Geoscience and Remote Sensing.
This personal contact can only
assist in the development of the
career of the students.
(a)
42
III. THE PROGRAMME
A. CONTACTS/ORGANIZERS
Xiuping Jia ([email protected]),
____________ The University of New
South Wales, Canberra Campus.
Kim Lowell ([email protected]),
Cooperative
______________
Research Centre for Spatial Information, University of
Melbourne.
Simon Jones ([email protected]),
_________________ RMIT University, Melbourne.
B. PROGRAMME
Thursday, July 18
9:00–9:10 AM
Opening
9:10–10:30
Nathan Quadros (Airborne LiDAR
Acquisition and Validation)
10:30–10:50
Morning Tea (provided)
10:50–12:10
Suzanne Furby (Image Processing
for the Land Cover Change Program
of a National Carbon Accounting
System)
(b)
(c)
FIGURE 1. Summer School 2013 organizers. (a) Xiuping Jia, (b) Kim Lowell, and
(c) Simon Jones.
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
12:10–1:30 PM
1:30–2:50
2:50–3:10
3:10–4:30
5:00–?
Friday, July 19
9:00–9:10 AM
9:10–10:30
10:30–10:50
10:50–12:10
12:10–1:30 PM
1:30–2:50
2:50–3:10
3:10–4:10
4:10–4:20
Lunch (on your own)
Mihai Datcu (Image Information
Mining)
Afternoon Tea (provided)
Lorenzo Bruzzone (Multi-Temporal
Image Analysis)
Social event
Housekeeping
Jocelyn Channusot (Hyperspectral
Image Analysis)
Morning Tea (provided)
Phil Tickle (The Challenges of Developing Australia’s High Resolution Coastal
DEM for Sea Level Rise and Coastal
Flood Applications)
Lunch (on your own)
John Richards (Thematic Mapping from
SAR Image Data)
Afternoon Tea (provided)
Mihai Tanase and Rocco Panciera (Calibrating airborne radar for soil-moisture
extraction)
Issue of certificate, photo, and wrap-up.
C. THE INSTRUCTORS
The GR4S in Melbourne benefitted from the following well
known instructors.
1) John Richards: John is a Fellow of the Australian Academy of Technological Sciences and Engineering, a Fellow
of the Institution of Engineers Australia and a Life Fellow
of the IEEE.
He has held the positions of Master of University House
at the Australian National University, Deputy Vice-Chancellor and Vice President of the ANU, and Dean of the Col-
lege of Engineering and Computer Science. In the 1980s he
was foundation Director of the Centre for Remote Sensing
at the University of New South Wales. John is an Emeritus Professor of the Australian National University and the
University of New South Wales, Visiting Professor at the
Harbin Institute of Technology, China, Visiting Professor
of the Chinese Academy of Sciences, and President of the
International Society for Digital Earth.
2) Lorenzo Bruzzone: received the Laurea (M.S.) degree
in electronic engineering (summa cum laude) and the
Ph.D. degree in telecommunications from the University of
Genoa, Italy, in 1993 and 1998, respectively.
He is currently a Full Professor of telecommunications
at the University of Trento, Italy, where he teaches remote
sensing, radar, pattern recognition, and electrical communications. Dr. Bruzzone is the founder and the director
of the Remote Sensing Laboratory in the Department of
Information Engineering and Computer Science, University of Trento. His current research interests are in the areas
of remote sensing, radar and SAR, signal processing, and
pattern recognition. He promotes and supervises research
on these topics within the frameworks of many national
and international projects. Among the others, he is the
Principal Investigator of the Radar for icy Moon exploration (RIME) instrument in the framework of the JUpiter
ICy moons Explorer (JUICE) mission of the European
Space Agency. He is the author (or coauthor) of 137 scientific publications in referred international journals (93 in
IEEE journals), more than 190 papers in conference proceedings, and 16 book chapters. He was invited as keynote
speaker in 24 international conferences and workshops.
He is a member of the Managing Committee of the Italian Inter-University Consortium on Telecommunications.
Since 2009 he is a member of the Administrative Committee of the IEEE Geoscience and Remote Sensing Society. He
is an IEEE Fellow.
FIGURE 2. GR4S Melbourne 2013.
SEPTEMBER 2013
43
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
(a)
(b)
(c)
(d)
FIGURE 3. Some scenes from the GR4S Melbourne. (a) Organizers and helpers: from left—Simon Jones, Xiuping Jia, Mariela Soto-Berelov
and Lola Suarez Barranco (Kim Lowell was unavailable). (b) The fun part of summer school in winter: relaxing at the bar. (c) Talking with the
instructor. (d) John Richards in action.
3) Mihai A. Tanase: received the engineering diploma in
forestry from Stefan cel Mare University,Romania, in 1999;
the diploma in economy from the Bucharest Academy of
Economic Studies, Romania, in 2004; the M.Sc. degree
in environmental management from the International
Centre for Advanced Mediterranean Agronomic Studies,
France, in 2007; and the Ph.D. degree in geography from
the University of Zaragoza, Spain, in 2010.
Since 2011, he holds a post doctoral position at the University of Melbourne/CRC for Spatial Information through
the ARC Super Science Fellowship program. He is a Principal Investigator for ESA-ERS, DLR-TerraSAR/Tandem-X,
JAXA-ALOS PALSAR2 and CSA-SOAR research programs
and participated in the JAXA K&C Initiative as a Coinvestigator. He is interested in the use of remote sensing for
vegetation characterization. His current activity is focused
on SAR, InSAR and PolSAR data for re severity estimation,
vegetation recovery monitoring and the retrieval of biogeophysical parameters of forests and agricultural crops.
Dr. Tanase was the recipient of the IEEE GRSS-IGARSS
Symposium Interactive Prize Paper Award in 2010.
4) Nathan Quadros: is an airborne LiDAR specialist
working at the Cooperative Research Centre for Spatial
Information. He has advised and managed the acquisition
of numerous LiDAR projects in Victoria, Australia and the
Pacific. His recent research has focused on vertical datum
44
transformations, LiDAR validation and acquisition standards. Nathan has a particular interest in the acquisition
of bathymetric LiDAR, and topographic LiDAR when it is
used for modelling sea-level rise, coastal inundation and
catchment flooding.
5) Rocco Panciera: received the M.S. degree in environmental engineering from the University of Trento, Italy, in
2003. In 2009 he obtained his Ph.D. degree in Environmental Engineering from the University of Melbourne, Australia,
with a thesis on the effect of land surface heterogeneity on
the accuracy of spaceborne soil moisture retrieval using passive microwave techniques. He then worked at the University
of Melbourne as a Research Fellow within the Soil Moisture
Active Passive (SMAPEx) project for algorithm development for NASAs SMAP mission. In this role he conducted
two large scale field experiments in South-Eastern Australia
including airborne active and passive microwave sensors.
From January 2011 he has been based at the Cooperative
Research Centre for Spatial Information (CRC-SI), Melbourne, Australia, where he is employed under the Superscience Fellowship from the Australian Research Council
(ARC). In this role he has been working on techniques to
estimate fine-resolution soil moisture by use of L-band and
X-band Synthetic Aperture Radar (SAR) measurements. He
has also been investigating the use of Light Detection and
Ranging (LiDAR) techniques for fine-resolution retrieval of
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
surface roughness characteristics in agricultural areas. His
current field of interest is the retrieval of soil moisture from
airborne and spaceborne SAR observations.
6) Suzanne Furby: is a member of the Terrestrial Mapping and Monitoring research group in CSIRO Mathematics, Informatics and Statistics, based in Perth, Western
Australia. She coordinated the remote sensing program to
monitor land cover change across the continent used in the
National Carbon Accounting System. The group continues
to work with the Department of Industry, Innovation, Climate Change, Science, Research and Tertiary Education to
develop further land cover change products for the NCAS
and for natural resource management applications more
generally. Suzanne is a key member of the team that is supporting Indonesia to build a National Carbon Accounting
System similar to that of Australia and has worked with
demonstration teams in China.
7) Jocelyn Chanussot (M04-SM04-F’12): received the M.Sc.
degree in electrical engineering from the Grenoble Institute of Technology (Grenoble INP), Grenoble, France, in
1995, and the Ph.D. degree from Savoie University, Annecy,
France, in 1998. In 1999, he was with the Geography Imagery Perception Laboratory for the Delegation Generale de
l’Armement (DGA-French National Defense Department).
Since 1999, he has been with Grenoble INP, where he was
an Assistant Professor from 1999 to 2005, an Associate Professor from 2005 to 2007, and is currently a Professor of
signal and image processing. He is conducting his research
at the Grenoble Images Speech Signals and Automatics Laboratory (GIPSA-Lab). His research interests include image
analysis, multicomponent image processing, nonlinear filtering, and data fusion in remote sensing.
Dr. Chanussot is the founding President of IEEE Geoscience and Remote Sensing French chapter (2007–2010) which
received the 2010 IEEE GRS-S Chapter Excellence Award. He
was the corecipient of the NORSIG 2006 Best Student Paper
Award, the IEEE GRSS 2011 Symposium Best Paper Award,
the IEEE GRSS 2012 Transactions Prize Paper Award and
the IEEE GRSS 2013 Highest Impact Paper Award. He was a
member of the IEEE Geoscience and Remote Sensing Society AdCom (2009–2010), in charge of membership development. He was the General Chair of the first IEEE GRSS
Workshop on Hyperspectral Image and Signal Processing,
Evolution in Remote sensing (WHISPERS). He was the Chair
(2009–2011) and Cochair of the GRS Data Fusion Technical
Committee (2005–2008). He was a member of the Machine
Learning for Signal Processing Technical Committee of the
IEEE Signal Processing Society (2006–2008) and the Program Chair of the IEEE International Workshop on Machine
Learning for Signal Processing, (2009). He was an Associate Editor for the IEEE Geoscience and Remote Sensing Letters
(2005–2007) and for Pattern Recognition (2006–2008). Since
2007, he is an Associate Editor for the IEEE Transactions on
Geoscience and Remote Sensing. Since 2011, he is the Editorin-Chief of the IEEE Journal of Selected Topics in Applied Earth
Observations and Remote Sensing.
IV. CONCLUSION
Once again, many thanks to the organizers of the GR4S
in Melbourne, and to the instructors who gave so freely of
their time and expertise. The opinions of the lectures by the
students was very positive, and the comments will be used
to fine tune the format of GR4S in future venues. If you feel
that your GRSS event might benefit from a Summer School,
please contact me.
I would also like to remind the community of our
quest for recently published Ph.D. theses. For publishing the Ph.D. thesis information you can contact Michael
Inggs ([email protected])
______________ or Dr. Lorenzo Bruzzone
([email protected]).
_______________ Ph.D. dissertations should be in
the fields of activity of IEEE GRSS and should be recently
completed. Please provide us with the following: title of
the dissertation, the students and advisors names, the
date of the thesis defense or publication, and a link for
downloading the electronic version of the thesis.
GRS
_____________________
___________
SEPTEMBER 2013
45
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
WOMEN IN GRS
FLORENCE TUPIN,
Telecom ParisTech, Paris, France
of data that has been acquired (an ocean of bytes every
day!), this link has become a key point in data enhancement, information extraction and data mining. In the
past, data were rare: a Ph.D. could be defended with
methods that had been applied to … a single image!
Nowadays, data are plenty, available through web sites
and Opportunity Announcements of space agencies.
The real difficulty now is how to handle all these data.
The most recent advances in image processing rely
both on more accurate mathematical modeling and on
computational power. This combined with huge data
set availability, has led to the emergence of new image
models. A striking example of this is in the so-called
“non-local” approaches for denoising or classification.
The main idea behind these methods is redundancy
which is by far a weaker assumption for an image than
regularity, sparsity of gradients, or smoothness and
texture decomposition hypotheses. This weak model,
This month’s Women in Geoscience
and Remote Sensing Magazine features
our first Guest Columnist: Florence
Tupin, Professor at Telecom ParisTech,
Paris, France. Please send comments
and suggestions for future articles or
guest columnists to gailsjackson@
________
ieee.org.
A
lthough I have written quite a few scientific
papers, I am a little undecided as to what to write
for this Women’s Column. Should it be on Remote
Sensing? On careers for women? On some new interesting scientific projects? (Or on bit on all of these topics?) Having carte blanche gives one freedom but can
also plunge one into the void.
Let me start by talking about image
processing, my field of specialization. I
worked at first on medical data and fingerprint images, but am glad I joined
the exciting Remote Sensing world afterwards. Image processing is a link in the
long remote sensing chain that starts
from sensors and goes to the applications to observe the Earth. As such, the
situation is not always quite comfortable.
One is usually not a real expert about
sensors and acquisition systems, but not
on geoscience and applications either.
Nevertheless, simply consider the huge
place taken by data processing in the past
fifteen years. With the enormous amount
46
Florence Tupin
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
however, was needed to handle a high diversity of images
and computational power has made it possible.
In my opinion, a fabulous experience for a researcher in
image processing is to follow the birth of a new sensor. For
example, CNES, the French space agency, and NASA are
working on SWOT project, which is a future topographic
mission for surface water and ocean measurement. SWOT
will combine a Nadir looking altimeter with the KaRIN
instrument, a Ka-band synthetic aperture radar. ESA is
working on BIOMASS project to take measurements of forest biomass with P-band synthetic aperture polarimetric
radar. These projects are ambitious projects, relying on
teams of hundreds of people working over several years to
provide new measurements of the Earth. Image and data
processing are small parts of these huge programs but are
absolutely essential if one is to extract valuable information efficiently and reliably. New advances in image processing have produced more efficient tools to process these
data. But at the same time new data, very high resolution
images (in spatial, temporal and spectral dimensions), and
the combination of sensors have raised new challenges for
researchers. Once again using the example of non-local
approaches, their adaptation to interferometric and polari-
metric data has made it necessary to define a new framework based on statistical modeling and that can be used
for a much wider set of data.
But what about women in Geoscience and Remote Sensing? Well, the situation is quite paradoxical... On the one
hand, at a purely local level, things seem to be changing.
Among the 14 Ph.D. students I have supervised, 9 were
women. And during my career, I have not confronted
any constraints linked to my gender, except, perhaps, the
Unconscious Bias (my own unconscious bias?) that was
referred to in the previous issue of this magazine. But on
the other hand, when considered from a broader vantage
point, the percentage of women in scientific domains does
remain weak. Moreover, the distinction between boys and
girls in society seem to be on the increase—for clothes,
shoes, children’s toys, etc., this undoubtedly heightened
by advertising. This tendency is without a doubt, much
more greater now than in my own childhood. One of the
main problems in scientific areas is the small percentage of
women working there, not their abilities. My final message
perhaps should be in favor of more Legos (even pink Legos)
and fewer Barbies (or at least Computer Engineer Barbies)
for little girls!
GRS
IEEE Open Access
6OSFTUSJDUFEBDDFTTUPUPEBZTHSPVOECSFBLJOHSFTFBSDIWJBUIF*&&&Xplore®EJHJUBMMJCSBSZ
*&&&PčFSTBWBSJFUZPGPQFOBDDFTT0"
QVCMJDBUJPOT
t)ZCSJEKPVSOBMTLOPXOGPSUIFJSFTUBCMJTIFEJNQBDUGBDUPST
t/FXGVMMZPQFOBDDFTTKPVSOBMTJONBOZUFDIOJDBMBSFBT
t"NVMUJEJTDJQMJOBSZPQFOBDDFTTNFHBKPVSOBMTQBOOJOHBMM
*&&&mFMETPGJOUFSFTU
%JTDPWFSUPQRVBMJUZBSUJDMFT
DIPTFOCZUIF*&&&QFFSSFWJFX
TUBOEBSEPGFYDFMMFODF
Learn more about IEEE Open Access
www.ieee.org/open-access
SEPTEMBER 2013
47
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
CONFERENCE REPORTS
MARTTI HALLIKAINEN AND WERNER WIESBECK,
IEEE GRSS Awards Committee Co-Chairs
IGARSS in Melbourne July 21–26, 2013
GRSS Major Awards and Fellow Recognitions at the Plenary Session
I
GARSS 2013 was organized in Melbourne with over
1,360 participants from 66 countries, 1025 oral papers
and 448 poster presentations.
The Plenary Session was held in the morning of July
22 in the Auditorium of the Melbourne Convention
and Exhibition Centre (MCEC), starting with a special
musical welcome. A more formal welcome was given by
General Cochairs Simon Jones and Peter Woodgate, followed by the Australian government’s video presentation by Senator the Honourable Kate Lundy, Minister
Assisting for Industry and Innovation, and responsible
for developing Australia’s satellite utilization program.
GRSS President Melba Crawford and IEEE President
Peter Staecker then presented GRSS and IEEE activities,
correspondingly.
Assistant Minister Kate Lundy recently released Australia’s Satellite Utilisation Policy. Under the policy, a
new Space Coordination Office was established on July
1, 2013, responsible for coordinating Australia’s domestic civilian space activities. The policy gives priority to
Earth observation from space, satellite communications, and positioning and navigation. Satellite imagery
alone was estimated in a 2010 report to contribute about
AUD 3.3 billion per year to the gross domestic product.
GRSS President Melba Crawford presented our Society’s activities and achievements. She showed Society
membership figures from 2000 to present, demonstrating an encouraging trend in spite of tough times
in global economy. The three GRSS journals—Transactions, Letters, and J-STARS—are constantly improving
impact factor with GRSS now reaching 3.47 and being
#2 in the field of remote sensing. All three journals are
now hybrid open access.
IEEE President Peter Crawford gave an overview
of the organization with convincing facts: IEEE has
48
425,000 members in more than 160 countries, more
than 116,000 student members, 45 Societies and Councils, 333 Sections in 10 geographic regions worldwide,
1300 sponsored conferences, 3 million documents in
IEEE Xplore digital library, and it publishes 150 transactions, journal and magazines.
Following these speeches GRSS Awards Cochair
Martti Hallikainen, IEEE President Peter Staecker, and
GRSS President Melba Crawford presented IEEE and
GRSS level awards and Fellow recognitions.
IEEE ELECTROMAGNETICS AWARD
The IEEE Electromagnetics Award was established by
the IEEE Board of Directors in 1996. It is sponsored
by IEEE Antennas and Propagation Society, IEEE Electromagnetic Compatibility Society, IEEE Microwave
Theory and Techniques Society, and the IEEE Geoscience and Remote Sensing Society. It is presented to recognize outstanding contributions to electromagnetics
in theory, application, or education. In the evaluation
process, the following criteria are considered: impact
on the profession, innovation/originality, quality of
publications, breadth, depth and duration of contributions, honors, other achievements, and the quality of
the nomination. The award consists of a bronze medal,
certificate, and honorarium.
The 2013 IEEE Electromagnetics Award is presented
to Prof. Leung Tsang from the University of Washington with the citation: “For contributions to statistical electromagnetic theory and applications in remote sensing of
complex geophysical environments.”
Leung Tsang (S’73–M’75–SM’85–F’90) was born
in Hong Kong. He received the S.B., S.M., and Ph.D.
degrees from the Department of Electrical Engineering and Computer Science of the Massachusetts Institute of Technology. He was with Schlumberger-Doll
Research Center in 1976–1978. He was on the faculty
of Texas A&M University in 1980–1983. Presently,
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Geoscience and Remote Sensing Society in 2000, the Distinguished Achievement Award from the IEEE Geoscience
and Remote Sensing Society in 2008 and the Fiorino Oro
Award from the CeTeM, Italy in 2010. He received the William Pecora Award cosponsored by the U.S. Department
of Interior and NASA in 2012. He is the recipient of the
2013 IEEE Electromagnetics Award. He has been member
of the Washington State Academy of Sciences since 2012.
FIGURE 1. General Cochairs Simon Jones and Peter Woodgate
opening IGARSS 2013.
he is a Professor of Electrical Engineering at the University of Washington, Seattle, Washington, where he has
taught since 1983. He was the Chair of the Department
from August 2006 to September 2011. Between 2001 and
2004, he was on leave with the Department of Electronic
Engineering of the City University of Hong Kong. He was
a Visiting Professor of the National Central University of
Taiwan between January to June 2013. Leung Tsang is the
coauthor of 4 books, Theory of Microwave Remote Sensing
and Scattering of Electromagnetic Waves Volumes 1, 2 and
3. His current research interests are in waves in random
media and rough surfaces, remote sensing and geoscience applications, computational electromagnetics, signal
integrity and optics.
He was the President of the IEEE Geoscience and
Remote Sensing Society in 2006–2007, and the Editorin-Chief of the IEEE Transactions on Geoscience and Remote
Sensing in 1996–2001. He was the Chair of the IEEE TAB
Periodicals Committee in 2008–2009, and the Chair of
the IEEE TAB Periodicals Review and Advisory Committee (PRAC) in 2010–2011. He is presently a Member-atLarge of the IEEE PSPB and a Member of PRAC.
He is a Fellow of IEEE and a Fellow the Optical Society
of America. He received the IEEE Third Millennium Medal
in 2000, the Outstanding Service Award from the IEEE
IEEE FELLOW RECOGNITIONS
The grade of IEEE Fellow recognizes unusual distinction
in the profession and shall be conferred only by invitation of the IEEE Board of Directors upon a person of
outstanding and extraordinary qualifications and experience in IEEE designated fields. The IEEE Bylaws limit
the number of members who can be advanced to Fellow
grade in any one year to 1‰ of the Institute membership, exclusive of students and affiliates. To qualify, the
candidate must be a Senior Member and be nominated
by an individual familiar with the candidate’s achievements. Endorsements are required from at least five IEEE
Fellows and an IEEE Society best qualified to judge. The
IEEE Fellow Committee, comprising about 50 IEEE Fellows, carefully evaluates all nominations and presents
a list of recommended candidates to the IEEE Board of
Directors for the final election.
The following GRSS members were elevated to the Fellow status effective January 1, 2013:
◗ Dr. Maurice Borgeaud from the European Space Agency,
Frascati, Italy
◗ Dr. Om-Prakash Narayan Calla from the International
Centre for Radio Science, Jodhpur, Rajasthan, India
◗ Dr. Mihai Datcu from the German Aerospace Center,
Oberpfaffenhofen, Germany
◗ Prof. Giles M. Foody from the University of Nottingham, UK
◗ Prof. Paolo Gamba from the University of Pavia, Italy
◗ Prof. Akira Hirose from the University of Tokyo, Japan
FIGURE 2. GRSS President Melba
FIGURE 3. IEEE President Peter Staecker
Crawford welcoming attendees and
presenting Society activities and
achievements.
providing an overview of IEEE.
SEPTEMBER 2013
FIGURE 4. Leung Tsang saying a few
words after he received the IEEE Electromagnetics Award from IEEE President
Peter Staecker. Note the bronze medal.
49
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
◗ Dr. Yann Kerr from CESBIO, Toulouse, France
◗ Dr. Gerhard Krieger from the German Aerospace Cen-
ter, Oberpfaffenhofen, Germany
◗ Dr. Riccardo Lanari from the CNR, Naples, Italy
◗ Prof. Shunlin Liang from University of Maryland, Col-
lege Park, MD, U.S.A.
◗ Prof. Christian Pichot from Universite de Nice-Sophia
Antipolis, France
Six out of the eleven new Fellows attended the Plenary
for their recognition: Dr. Mihail Datcu, Prof. Paolo Gamba,
Prof. Akira Hirose, Dr. Yann Kerr, Prof. Shunlin Liang, and
Prof. Giles M. Foody, The others were recognized with a
round of applause.
Dr. Mihai Datcu was elevated to Fellow grade with the
citation: “For contributions to information mining of high resolution synthetic aperture radar and optical Earth observation images.”
Mihai Datcu received the M.S. and Ph.D. degrees in
electronics and telecommunications from the University
Politehnica of Bucharest (UPB), Romania, in 1978 and
1986, respectively. In 1999, he received the title “Habilitation à diriger des recherches” in computer science from
University Louis Pasteur, Strasbourg, France. Since 1981,
he has been a Professor with the Faculty of Electronics,
Telecommunications and Information Technology, UPB,
working in signal/image processing and Electronic Speckle
Interferometry. Since 1993, he has been a scientist with
the German Aerospace Center (DLR), Oberpfaffenhofen,
Germany. He is developing algorithms for analyzing Very
High Resolution Synthetic Aperture Radar (VHR SAR)
and Interferometric SAR (InSAR) data. He is engaged in
research related to information theoretical aspects and
semantic representations in advanced communication systems. Currently, he is Senior Scientist and Image Analysis
research group leader with the Remote Sensing Technology Institute of DLR, Oberpfaffenhofen. Since 2011, he
is also leading the Immersive Visual Information Mining
research laboratory at the Munich Aerospace Faculty and
is director of the Research Center for Spatial Information
at UPB. He has held Visiting Professor appointments with
the University of Oviedo, Spain, University Louis Pasteur
and the International Space University, in Strasbourg,
France, University of Siegen, Germany, University of Camerino, Italy, and the Swiss Center for Scientific Computing, Manno, Switzerland. From 1992 to 2002 he had a
longer Invited Professor assignment with the Swiss Federal
Institute of Technology, ETH Zurich. Since 2001, he has
initiated and led the Competence Centre on Information
Extraction and Image Understanding for Earth Observation, at ParisTech, Telecom Paris, a collaboration of DLR
with the French Space Agency (CNES). He has been Professor holder of the DLR-CNES Chair at ParisTech, Telecom Paris. His interests are in information and complexity theory, stochastic processes, Bayesian inference, and
Image Information Mining (IIM). He and his team have
developed and are currently developing the operational
IIM processor in the Payload Ground Segment systems
50
FIGURE 5. Mihai Datcu (right) receives his Fellow recognition from
IEEE President Peter Staecker.
FIGURE 6. Paolo Gamba (right) receives his Fellow recognition
from IEEE President Peter Staecker.
for the German missions TerraSAR-X, TanDEM-X, and
the ESA Sentinel 1 and 2. He is the author of more than
200 scientific publications, among them about 50 journal
papers, and a book on number theory. He is a member of
the European Image Information Mining Coordination
Group (IIMCG) and of the Data Archiving and Distribution Technical Committee (DAD TC) of the IEEE Geoscience and Remote Sensing Society.
Prof. Paolo Gamba was elevated to Fellow grade with
the citation: “For contributions to very high resolution remote
sensing image processing of urban areas.”
Paolo Gamba (SM’00, F’13) is currently Associate Professor of Telecommunications at the University of Pavia,
Italy. He received the Laurea degree in Electronic Engineering “cum laude” from the University of Pavia, Italy, in 1989,
and the Ph.D. degree in Electronic Engineering from the
same University in 1993. He is a Fellow of IEEE, and since
January 2009 he has been serving as Editor-in-Chief of the
IEEE Geoscience and Remote Sensing Letters. He has been the
organizer and Technical Chair of the biennial GRSS/ISPRS
Joint Workshops on “Remote Sensing and Data Fusion over
Urban Areas” since 2001. The next conference in the series,
JURSE 2013, is going to be Sao Paulo in April 2013. He also
served as Technical Co-chair of the 2010 IEEE Geoscience
and Remote Sensing Symposium, Honolulu, Hawaii, July
2010, and will serve as Technical Cochair of the 2015 IEEE
Geoscience and Remote Sensing Symposium, in Milan,
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
FIGURE 7. Akira Hirose (right) receives his Fellow recognition from
FIGURE 8. Yann Kerr (right) receives his Fellow recognition from
Italy. He has been Chair of Technical Committee 7 “Pattern Recognition in Remote Sensing” of the International
Association for Pattern Recognition (IAPR) from October
2002 to October 2004 and Chair of the Data Fusion Committee of the IEEE Geoscience and Remote Sensing Society
from October 2005 to May 2009. He has been the Guest
Editor of special issues of IEEE Transactions on Geoscience and
Remote Sensing, IEEE Journal of Selected Topics in Remote Sensing Applications, ISPRS Journal of Photogrammetry and Remote
Sensing, International Journal of Information Fusion and Pattern
Recognition Letters on the topic of Urban Remote Sensing,
Remote Sensing for Disaster Management, Pattern Recognition in Remote Sensing Applications. He has been invited
to give keynote lectures and tutorials in several occasions.
He has published 100 papers in international peer-review
journals and presented nearly 250 research works in workshops and conferences.
Prof. Akira Hirose was elevated to Fellow grade with
the citation: “For contributions to theory and radar applications
of complex-valued neural networks.”
Akira Hirose received the Ph.D. degree from the University of Tokyo, Tokyo, Japan, in 1991 in electronic engineering. In 1987, he joined the Research Center for Advanced
Science and Technology (RCAST), The University of
Tokyo, as a Research Associate. In 1991, he was appointed
an Instructor at the RCAST. From 1993 to 1995, on leave of
absence from the University of Tokyo, he joined the Institute for Neuroinformatics, University of Bonn, Germany.
SEPTEMBER 2013
Presently he is a Professor at the Department of Electrical
Engineering and Information Systems, The University of
Tokyo, Japan. The main fields of his interest are wireless
electronics and neural networks. He served as the Editorin-Chief of the IEICE Transactions on Electronics, Associate
Editor of journals such as the IEEE Transactions on Neural
Networks, and the Chair of the Neurocomputing Technical Group in the Institute of Electronics, Information and
Communication Engineers (IEICE). He serves presently as
the Vice-President of Japanese Neural Networks Society
(JNNS), Member of IEEE CIS Neural Networks Technical Committee, Governing Board Member of Asia-Pacific
Neural Network Assembly (APNNA), the General Chair
of Asia-Pacific Conference on Synthetic Aperture Radar
(APSAR) 2013 Tsukuba and Associate Editor of the IEEE
Geoscience and Remote Sensing Newsletter. Dr. Hirose is a
Senior Member of the IEICE and a Member of the JNNS.
Dr. Yann Kerr was elevated to Fellow grade with the
citation: “For contributions to two-dimensional L-band microwave interferometer design.”
Yann H. Kerr (M ’88, SM ’01, F’13) received the engineering degree from Ecole Nationale Supérieure de
l’Aéronautique et de l’Espace, the M.Sc. degree in electronics and electrical engineering from Glasgow University,
Glasgow, Scotland, UK, and the Ph.D. degree in Astrophysique Gophysique et Techniques Spatiales, Université
Paul Sabatier, Toulouse, France. His fields of interest are
in the theory and techniques for microwave and thermal
infrared remote sensing of the Earth, with emphasis on
hydrology, water resources management. He is currently
the Director of Centre d’Etudes Spatiales de la BIOsphère,
Toulouse, France. He was an EOS principal investigator
(interdisciplinary investigations) and PI and precursor
of the use of the SCAT over land. In 1990 he started to
work on the interferometric concept applied to passive
microwave Earth observation and was subsequently the
science lead on the MIRAS project for ESA. In 1997 he proposed the SMOS Mission, the natural outcome of the previous MIRAS work. He is currently involved in the exploitation of SMOS data, in the Cal Val activities and related
level 2 soil moisture and level 3 and 4 developments. He
is also working on the SMOS next concept and involved
in both the Aquarius and SMAP missions. He received the
World Meteorological Organization 1st prize (Norbert
Gerbier), the USDA Secretary’s team award for excellence
(Salsa Program), the GRSS certificate of recognition for
leadership in development of the first synthetic aperture
microwave radiometer in space and success of the SMOS
mission, and is a Distinguished Lecturer for GRSS.
Prof. Shunlin Liang was elevated to Fellow grade with
the citation: ”For contributions to remote sensing from satellite
observations.”
Shunlin Liang (M’94-SM’01-F’13) received his Ph.D.
degree in 1993 from Boston University, Boston, MA, USA. He
is currently a Professor with the department of geographical
sciences, University of Maryland, College Park, MD, USA,
51
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
FIGURE 9. Shunlin Liang (right) receives his Fellow recognition
from IEEE President Peter Staecker.
FIGURE 10. Giles Foody (second from right) flew to Melbourne on
Monday morning, but made it to the Plenary to receive his Fellow
recognition from GRSS President Melba Crawford (second from left).
IEEE President Peter Staecker (right) and GRSS Awards Cochair Martti
Hallikainen (left) joined Melba in congratulating him.
and the college of global change and earth system sciences,
Beijing Normal University, Beijing, China. His main research
interests focus on estimation of land surface variables from
satellite data, earth energy balance, and assessment of environmental impacts of vegetation changes. He published about
160 peer-reviewed journal papers, authored the book Quantitative Remote Sensing of Land Surfaces (Wiley, 2004), edited the
book Advances in Land Remote Sensing: System, Modeling, Inversion and Application (Springer, 2008), and co-edited the book
Advanced Remote Sensing: Terrestrial Information Extraction and
Applications (Academic Press, 2012). He is an Associate Editor
of the IEEE Transactions on Geoscience and Remote Sensing and
also a guest editor of multiple remote sensing journals.
Prof. Giles M. Foody was elevated to Fellow grade with
the citation: “For contributions to the remote sensing of land cover.”
Giles M. Foody (M’01, SM’11, F’13) earned the B.Sc.
and Ph.D. degrees from the University of Sheffield, U.K.,
in 1983 and 1986, respectively. He is currently Professor
of Geographical Information Science at the University of
Nottingham, U.K. His main research interests focus on the
interface between remote sensing, ecology, and informatics. Professor Foody is currently the Editor-in-Chief of the
International Journal of Remote Sensing and of the recently
launched journal Remote Sensing Letters. He holds editorial
roles with Landscape Ecology and Ecological Informatics
and serves on the editorial board of several other journals.
The following five new Fellows were unfortunately not
able to attend the Plenary.
Dr. Maurice Borgeaud was elevated to Fellow grade
with the citation: “For leadership in microwave remote sensing
from spaceborne systems and retrieval of bio-physical and geophysical parameters for land applications.”
Maurice Borgeaud earned a “Diplôme d’Ingénieur Electricien” from EPFL in Lausanne and then received a Master
of Science and a Ph.D. Degree (1987) from the Massachusetts Institute of Technology, both in Electrical Engineering
and Computer Science. He worked for the German Aerospace Agency (DLR) in Munich and was then member of
the staff of the European Space Agency (ESA) from 1989 to
2002 at the European Space Research and Technology Center (ESTEC). During this time, he was responsible for several
space technology and Earth observation projects. He also
spent a one-year sabbatical leave at the NASA/CALTECH Jet
FIGURE 11. Five new Fellows at the Plenary (from left): Yann Kerr, Shunlin Liang, GRSS Awards Cochair Martti Hallikainen, GRSS President
Melba Crawford, IEEE President Peter Staecker, Akira Hirose, Paolo Gamba, and Mihai Datcu. Giles Foody made it to the Plenary a few
moments later.
52
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Gerhard Krieger
Riccardo Lanari
Christian Pichot
Maurice Borgeaud
O.P.N. Calla
FIGURE 12. Five new Fellows could not make it to the Plenary.
Propulsion Laboratory (JPL) in 1992 and received a Master
of Space Systems Engineering from TU-Delft in 1997. From
2002 to 2004, he was affiliated with the Swiss Space Office
and was member of the Swiss delegation to ESA where he
was in charge of the Earth observation, satellite navigation
and communications sectors. Between 2004 and 2010, he
was the Director of the Space Center EPFL where he successfully led the university team which developed and built the
first Swiss student Cubesat satellite named “SwissCube”.
Between 2008 and 2010, he was Chairman of the ESA Programme Board on Earth observation.
Since 2011, Maurice Borgeaud is back with ESA as Head
of the department dealing with science, applications, and
future technologies in the Earth Observation Directorate.
His main tasks are to interact with both the scientific community and industry in order to propose new innovative
spaceborne missions. He also participated in the recent definition of the long-term EO exploitation strategy to be realized via a coherent programmatic framework that addresses
the full spectrum of EO user communities, including science, applications and services development. He is also
managing the ESA effort on climate change and the development of essential climate variables. Dr. Borgeaud is a
Member of the IEEE since 1979 and is an Associate Editor
for IEEE Transactions on Geoscience and Remote Sensing. He
has authored and coauthored more than 100 publications
in refereed journals or conference proceedings.
Dr. Om-Prakash Narayan Calla was elevated to Fellow
grade with the citation: “For leadership in space applications of
microwave technology and remote sensing.”
O.P.N. Calla, FIEEE, FNAE, Dist. FIETE, FIE, FGAS,
FVEDA, MGBMS, MISRS, MIMS, MASI, MATMS, FBES
did his Post graduation (ME) in Electronics from Birla
Engineering College, Pilani and was awarded with a silver medal. In December 1995, he retired as Scientist ‘H’
and Deputy Director, Satellite Communication Area of
Space Application Centre (SAC), ISRO, Ahmedabad. He is
responsible to initiate Microwave Remote Sensing Activities in India. He is known as Father of Microwave Remote
Sensing in India. Presently, he is the Director of International Center for Radio Science (ICRS) from May 1996.
After completing his Masters, Prof. Calla was involved in
research and development work at Atomic Energy EstabSEPTEMBER 2013
lishment, Trombay and INCOSPAR from 1962–70. From
1962–1963 he was involved in the RADAR Development at
AEET (BARC). He made radar transponders during 1964–
1969. The First Rocket Borne Microwave Payload in India
was also developed by him during this time only. In 1970,
he was appointed as Head, Electronics Division of Microwave Antenna Systems Engineering Group (MASEG), and
Secretary MASEG. At MID, he developed Low Cost Antenna
for Rural TV for Site Experiment. He also developed feed
for satellite communication earth station. In 1972, he was
appointed as Head of Microwave Division, ISRO and was
responsible for development of dual polarized feed for
troposcatter and satellite communication antenna feeds.
From 1977–85 he worked as Chairman, Communications
Area of Space Applications Centre (SAC) of Indian Space
Research Organisation (ISRO) and as Principal Scientist
and Responsible for Development of Satellite Microwave
Radiometer (SAMIR), one of the primary payloads of
BHASKARA-I & II Satellites. Because of him, India became
the third country in world who has put microwave remote
sensing payload onboard a satellite. In 1985–88 he designated as Programme Director, Microwave Remote Sensing Programme of Indian Space Research Organization
(ISRO). He was responsible for development of Airborne
SLAR.
After his retirement he established ICRS at Jodhpur, a
centre to carry on MRS activities in India. At ICRS, he has
been a principal investigator of more than 18 R&D projects related to microwaves and their applications. At ICRS,
in the last few years the work related to measurement of
electrical properties of natural earth materials, lossy and
lossless materials used for missiles and the terrestrial Analogue of Lunar soil is being done. He has been working on
soil moisture mapping over India & ocean salinity retrieval
using MRS techniques. He is also working on physical and
electrical modeling of the Lunar Surface. He is the first
Indian who has developed microwave hardware for all the
platforms viz. ground, air, rocket and space borne.
He was honored with Hari Om Ashram Award by PRL
Ahmedabad in 1978 for the contribution in the field of Satellite Communications, Ram Lal Wadhwa Gold Medal in
1979 by IETE, Delhi for contribution in the field of Electronics and Communications for last 10 years, President
53
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
of India’s Prize (Hindi) for 1984–85 for the paper titled
“Modern Maritime Satellite Communication System—
INMARSAT” which was published in the Journal of Institution of Engineers (India), Calcutta Vol.65, Part Hindi
3, April 1985, Consolation Prize for Hindi Book titled
Antariksh Mein Bharat under the Indira Gandhi Rajbhasha
Scheme for the year 1986–87 for writing original Hindi
book, Certificate of Merit for 1992–93 by Institution of
Engineers (India), Calcutta for the paper entitled “Meteorological Satellites” published in their journal and by now
he has published more than 500 papers in various national
& international journals and conferences. He has been
conferred Marwar Ratna Award for year 2007 by Mehrangarh Museum Trust, Jodhpur. He also served as President
of IETE in the year 1991. He was Chairman of Ahmedabad
Chapter of IETE, CSI, ISRS, IMS and GBMS. Recently he
has received Lifetime Achievement Award from VEDA. He
was awarded by ISM, the ISM-2010 Pioneer in Microwave
Academics Award in 2010, and was Conferred ISM 2010
Pioneer in Microwave Engineering.
Dr. Gerhard Krieger was elevated to Fellow grade with
the citation: “For contributions to advanced synthetic aperture
radar systems.”
Gerhard Krieger received the Dipl.-Ing. (M.S.) and Dr.Ing. (Ph.D.) degrees (with honors) in electrical and communication engineering from the Technical University of
Munich, Germany, in 1992 and 1999, respectively. From
1992 to 1999, he was with the Ludwig Maximilians University, Munich, where he conducted multidisciplinary
research on neuronal modeling and nonlinear information processing in biological and technical vision systems.
Since 1999, he has been with the Microwaves and Radar
Institute (HR) of the German Aerospace Center (DLR),
Oberpfaffenhofen, Germany, where he developed signal and image processing algorithms for a novel forward
looking radar system employing digital beamforming
on receive. From 2001 to 2007 he led the New Synthetic
Aperture Radar (SAR) Missions Group which pioneered
the development of advanced bistatic and multistatic
radar systems as exemplified by the TanDEM-X mission,
as well as innovative multichannel SAR techniques and
algorithms for high resolution wide-swath SAR imaging.
Since 2008, he has been the Head of the new Radar Concepts Department of the Microwaves and Radar Institute,
DLR, Oberpfaffenhofen, Germany. Gerhard Krieger has
authored more than 50 peer-reviewed journal papers, 7
invited book chapters, about 300 conference papers, and
5 patent families. His current research interests focus on
the development of multichannel radar techniques and
algorithms for innovative Multiple-Input Multiple-Output
(MIMO) SAR systems, the demonstration of novel interferometric and tomographic Earth observation applications,
and the conceptual design of advanced bi- and multistatic
radar missions. Dr. Krieger received several national and
international awards, including the W.R.G. Baker Prize
Paper Award from the IEEE Board of Directors and the
54
Transactions Prize Paper Award of the IEEE Geoscience
and Remote Sensing Society. In 2012, he and his colleagues were nominated for the German President’s Award
for Technology and Innovation. Since 2012, he has been
Associate Editor of the IEEE Transactions on Geoscience and
Remote Sensing.
Dr. Riccardo Lanari was elevated to Fellow grade with the
citation: “For contributions to synthetic aperture radar processing.”
Riccardo Lanari was born in Naples, Italy, in 1964. He
graduated in electronic engineering (summa cum laude)
from the University of Naples, Federico II, Naples, in 1989.
In the same year, following a short experience at ITALTEL SISTEMI SPA, he joined IRECE and after Istituto per
il Rilevamento Elettromagnetico dell’Ambiente (IREA), a
Research Institute of the Italian National Research Council
(CNR), Naples, where, since November 2011, he is the Institute Director. He has lectured in several national and foreign universities and research centers. He was an Adjunct
Professor of electrical communication with the l’Università
del Sannio, Benevento, Italy, from 2000 to 2003, and from
2000 to 2008, he was the Lecturer of the synthetic aperture radar (SAR) module course of the International Master
in Airbone Photogrammetry and Remote Sensing offered
by the Institute of Geomatics, Barcelona, Spain. He was a
Visiting Scientist at different foreign research institutes,
including the Institute of Space and Astronautical Science,
Japan, in 1993; German Aerospace Research Establishment (DLR), Germany, in 1991 and 1994; and Jet Propulsion Laboratory, Pasadena, CA, in 1997, 2004, and 2008.
His main research activities are in the SAR data processing
field as well as in SAR interferometry techniques; on this
topic, he is the holder of two patents, and he has authored
or coauthored more than 70 international journal papers
and the book Synthetic Aperture Radar Processing (1999,
CRC Press). Dr. Lanari is a Distinguished Speaker of the
Geoscience and Remote Sensing Society of IEEE, and he
has served as the chairman and as a technical program
committee member at several international conferences.
Moreover, he acts as a reviewer of several peer-reviewed
international journals. He received a NASA recognition
and a group award for the technical developments related
to the Shuttle Radar Topography Mission.
Prof. Christian Pichot was elevated to Fellow grade
with the citation: “For contributions to microwave tomography
and antenna designs.”
Christian Pichot is currently a Research Director at the
French National Center for Scientific Research (CNRS),
working at the Electronics, Antennas & Telecommunications Laboratory (LEAT), University Nice-Sopihia Antipolis and CNRS, Sophia Antipolis, France. He is also codirector of The Centre de REcherche Mutualisé sur les ANTennes
(CREMANT), a Joint Antenna Research Center, founded
in September 2008, supported by the University of NiceSophia Antipolis, CNRS and France Telecom Orange Labs.
He received the Ph.D. and the Doctor of Science (D.Sc.)
degrees from the University of Paris-Sud 11 in 1977 and
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
1982, respectively. In 1978, he joined the Systems and
Signals Laboratory CNRS/Supélec, Gif-sur-Yvette, France,
where he was the Waves Division Leader from 1991 to
1992. During the 1989–1990 academic year, he was a Visiting Researcher at the Lawrence Livermore National Laboratory, Livermore, CA, USA. From 1993 to 1995, he was with
the Sophia Antipolis Computer Science, Signals and Systems Laboratory, University Nice-Sophia Antipolis/CNRS,
Valbonne, France. In 1996, he joined Electronics, Antennas
& Telecommunications Laboratory (LEAT) and from 2000
to 2011, he was Director of LEAT. In 2006 he was the local
co-organizer of the 1st European Conference on Antennas
and Propagation (EuCAP 2006) in Nice, France. He has
organized and chaired numerous special sessions in International Symposiums and IEEE Conferences (including
AP-S conferences) and Workshops. He received in 1983 the
European Microwave Prize. He has been elected in 2012 as
a member of the IEEE AP-S AdCom. His research activities
are concerned with scattering and propagation of Electromagnetic Waves, radiation of antennas, inverse scattering
(Microwave Imaging and Tomography, Antenna Synthesis,
Complex Permittivity Reconstruction, Object Detection
and Recognition) for applications in Radar, Civil engineering, non-destructive evaluation (NDE), non-destructive
testing (NDT), geophysics engineering, security and military applications, antennas, telecommunications, and
medical domain (bio-engineering), VLF/LF frequencies,
microwaves and millimetre waves.
The Geoscience and Remote Sensing Society is proud to
have a total of eleven new Fellows, working in Italy, France,
Germany, the U.S., UK, India, and Japan. This is another
indication on the international nature of our Society.
IEEE GRSS GOLD EARLY CAREER AWARD
The GRSS GOLD Early Career Award is to promote,
recognize and support young scientists and engineers
within the Geoscience and Remote Sensing Society that
have demonstrated outstanding ability and promise for
significant contributions in the future. Selection factors
FIGURE 13. Carlos López-Martínez (right) exchanges a few words
with GRSS President Melba Crawford after receiving the GOLD Early
Career Award from her. IGARSS General Cochair Simon Jones is in
the background.
SEPTEMBER 2013
include quality, significance and impact of contributions,
papers published in archival journals—papers presented
at conferences and symposia, patents, demonstration of
leadership, and advancement of profession. The candidate must be an IEEE GRSS Graduate of the Last Decade
(GOLD) member at the time of nomination and making
contributions in a GRSS field of interest. Previous award
winners are ineligible. The Award consists of a Certificate
and honorarium.
The 2013 GOLD Early Career Award is presented to Dr.
Carlos López-Martínez from Universitat Politècnica de
Catalunya (UPC), Barcelona, Spain, with the citation “In
recognition of his outstanding ability and promise for significant
contributions in the future.”
Carlos Lopez-Martinez received the MSc. degree in
electrical engineering and the Ph.D. degree from the Universitat Politècnica de Catalunya, Barcelona, Spain, in
1999 and 2003, respectively. From October 2000 to March
2002, he was with the Institut für Hochfrequenztechnik
and Radar Systeme IHR, German Aerospace Center, DLR,
Oberpfaffenhofen, Germany. From June 2003 to December 2005, he has been with the Image and Remote Sensing Group—SAPHIR Team, in the Institute of Electronics
and Telecommunications of Rennes (I.E.T.R.—CNRS UMR
6164), Rennes, France. In January 2006, he joined the
Universitat Politècnica de Catalunya, Barcelona, Spain, as
a Ramón-y-Cajal researcher, where he is currently associate professor in the area of remote sensing and microwave
technology. His research interests include SAR and multidimensional SAR, radar polarimetry, physical parameter
inversion, digital signal processing, estimation theory and
harmonic analysis. He is associate editor of IEEE Journal of
Selected Topics in Applied Earth Observations and Remote Sensing and he served as guest editor of the EURASIP Journal
on Advances in Signal Processing. He has organized different invited sessions in international conferences on radar
and SAR polarimetry. He has presented advanced courses
and seminars on radar polarimetry to a wide range of organizations and events. Dr. López-Martínez has authored
or coauthored more than 100 articles in journals, books,
and conference proceedings in the Radar Remote Sensing and image analysis literature. He received the Student
Prize Paper Award at the EUSAR 2002 Conference and
coauthored the paper awarded with the First Place Student
Paper Award at the EUSAR 2012 Conference.
IEEE GRSS MAJOR AWARDS
The call for nominations for the GRSS Distinguished
Achievement Award, GRSS Outstanding Service Award
and the GRSS Education Award are published in the GRS
Magazine. The nomination forms are available on the GRSS
home page (http://www.grss-ieee.org/about/awards/).
Any member, with the exception of GRSS Administrative
Committee members, can make nominations to recognize deserving individuals. Typically the lists of candidates comprise three to five names. An independent Major
55
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Awards Committee makes the selection, which is approved
by the GRSS AdCom.
IEEE GRSS EDUCATION AWARD
The Education Award was established to recognize an individual who has made significant educational contributions
to the field of GRSS. Selection criteria include significance
of the educational contribution in terms of innovation
and the extent of its overall impact. The contribution can
be at any level, including K-12, undergraduate and graduate teaching, professional development, and public outreach. It can also be in any form (e.g. textbooks, curriculum development, educational program initiatives). IEEE
GRSS membership or affiliation is required. The awardee
receives a certificate.
The 2013 Education Award is presented to Prof. Kamal
Sarabandi from the University of Michigan, Ann Arbor, MI,
USA, with the citation: “In recognition of his significant educational contributions to Geoscience and Remote Sensing.”
Kamal Sarabandi (S’87-M’90-SM’92-F’00) received
the B.S. degree in electrical engineering from the Sharif
University of Technology, Tehran, Iran, in 1980, the M.S.
degree in electrical engineering in 1986, and the M.S.
degree in mathematics and the Ph.D. degree in electrical
engineering from The University of Michigan at Ann Arbor
in 1989.
He is currently the Director of the Radiation Laboratory and the Rufus S. Teesdale Professor of Engineering in
the Department of Electrical Engineering and Computer
Science, The University of Michigan at Ann Arbor. His
research areas of interest include microwave and millimeter-wave radar remote sensing, Meta-materials, electromagnetic wave propagation, and antenna miniaturization.
He possesses 25 years of experience with wave propagation in random media, communication channel modeling, microwave sensors, and radar systems and leads a
large research group including two research scientists, 16
Ph.D. students. He has graduated 40 Ph.D. and supervised
numerous post-doctoral students. He has served as the
Principal Investigator on many projects sponsored by the
National Aeronautics and Space Administration (NASA),
Jet Propulsion Laboratory (JPL), Army Research Office
(ARO), Office of Naval Research (ONR), Army Research
Laboratory (ARL), National Science Foundation (NSF),
Defense Advanced Research Projects Agency (DARPA),
and a large number of industries. Currently he is leading
the Center for Microelectronics and Sensors funded by the
Army Research Laboratory under the Micro-Autonomous
Systems and Technology (MAST) Collaborative Technology Alliance (CTA) program.
He has published many book chapters and more than
220 papers in refereed journals on miniaturized and onchip antennas, meta-materials, electromagnetic scattering, wireless channel modeling, random media modeling,
microwave measurement techniques, radar calibration,
inverse scattering problems, and microwave sensors. He
56
FIGURE 14. Kamal Sarabandi (right) receives the IEEE GRSS Educ-
tion Award from GRSS President Melba Crawford.
has also had more than 500 papers and invited presentations in many national and international conferences and
symposia on similar subjects.
Dr. Sarabandi served as a member of NASA Advisory
Council appointed by the NASA Administrator for two
consecutive terms from 2006–2010. He is serving as a vice
president of the IEEE Geoscience and Remote Sensing Society (GRSS) and is a member of the Editorial Board of the
Proceedings of the IEEE. He was an associate editor of the
IEEE Transactions on Antennas and Propagation and the IEEE
Sensors Journal. He is a member of Commissions F and D
of URSI. Dr. Sarabandi was the recipient of the Henry Russel Award from the Regent of The University of Michigan.
In 1999 he received a GAAC Distinguished Lecturer Award
from the German Federal Ministry for Education, Science,
and Technology. He was also a recipient of the 1996 EECS
Department Teaching Excellence Award and a 2004 College of Engineering Research Excellence Award. In 2005
he received the IEEE GRSS Distinguished Achievement
Award and the University of Michigan Faculty Recognition Award. He also received the best paper Award at the
2006 Army Science Conference. In 2008 he was awarded a
Humboldt Research Award from The Alexander von Humboldt Foundation of Germany and received the best paper
award at the IEEE Geoscience and Remote Sensing Symposium. He was also awarded the 2010 Distinguished Faculty
Achievement Award from the University of Michigan. The
IEEE Board of Directors announced him as the recipient of
the 2011 IEEE Judith A. Resnik medal. In the past several
years, joint papers presented by his students at a number
of international symposia (IEEE APS’95,’97,’00,’01,’03,
’05,’06,’07; IEEE IGARSS’99,’02,’07;’11 IEEE IMS’01, USNC
URSI’04,’05,’06,’10,’11 AMTA ’06, URSI GA 2008) have
received best paper awards.
IEEE GRSS OUTSTANDING SERVICE AWARD
The Outstanding Service Award is not presented in 2013.
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
FIGURE 15. Roger Lang saying a few words after receiving the
IEEE GRSS Distinguished Achievement Award from GRSS President
Melba Crawford. IEEE President Peter Staecker is listening.
IEEE GRSS DISTINGUISHED ACHIEVEMENT AWARD
The Distinguished Achievement Award was established to
recognize an individual who has made significant technical contributions, within the scope of GRSS, usually over
a sustained period. Selection criteria include quality, significance and impact of the contributions; quantity of
the contributions; duration of significant activity; papers
published in archival journals; papers presented at conferences and symposia; patents granted; and advancement
of the profession. IEEE membership is preferable but not
required. The award is considered annually and presented
only if a suitable candidate is identified. The awardee
receives a plaque and a certificate.
The 2013 IEEE GRSS Distinguished Achievement Award
is presented to Prof. Roger H. Lang from the George Washington University, Washington, D.C., USA with the citation: “For innovative and continuing contributions to the theory
of microwave models for remote sensing, propagation and measurements, and his service to the community.”
Roger H. Lang received his education at the Polytechnic Institute of Brooklyn (now Polytechnic Institute at
New York University) and earned a Ph.D. in Electrophysics. He did post-doctoral work on wave propagation in
random medium at the Courant Institute of Mathematical Sciences, New York University under JB Keller, and
then joined the George Washington University where
he became a Professor and was awarded the L. Stanley
Crane Chair.
Roger Lang’s research has focused on microwave remote
sensing and wave propagation, and especially the effect of
vegetation on remote sensing. He was the first to employ
the Distorted Born Approximation, in conjunction with
the discrete scatter model, to describe the effects of a vegetation canopy. This was significant because it was fully
consistent with Maxwell’s equations and related the scattering and propagation to physical quantities of the canopy (e.g. the biophysical characteristics of leaves, stems
and branches). The theory provided a systematic treatment
of the double bounce contribution from vegetation which
is important in the measurement of soil moisture and vegetation biomass. In 1989 Professor Lang was elected a Fellow of the IEEE based in large part on this work.
Dr. Lang along with his students and collaborators
applied the Distorted Born methodology in numerous
NASA field campaigns. Many of these measurements
involved a radar/ radiometer now known as ComRAD that
he built in conjunction with NASA/GSFC. In all of these
campaigns careful biophysical measurements of vegetation were made to help understand how models and measurements compared. Dr. Lang has made dielectric measurements of vegetation components and more recently
has been involved in making measurements of the dielectric constant of seawater for the Aquarius project.
Dr. Lang has been an active participate in IEEE GRSS.
He has served on the GRSS AdCom and has been an
associate editor of TGARS for many years. He served as
Co-Technical Chair of IGARSS 1991, and more recently
was the Chair of MicroRad 2010. Dr. Lang is also active
in URSI where he is presently the Chair of International
URSI Commission F. He has over 200 journal and conference papers.
FIGURE 16. New Fellows and award recipients (from left): Carlos López-Martínez, Paolo Gamba, Yann Kerr, Shunlin Liang, GRSS Awards Cochair
Martti Hallikainen, GRSS President Melba Crawford, IEEE President Peter Staecker, Leung Tsang, Akira Hirose, Mihai Datcu, Kamal Sarabandi, and
Roger Lang.
SEPTEMBER 2013
57
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
FIGURE 17. Dr. Chris Pigram discussing geoscience and remote sensing in
Australia.
FIGURE 18. The topic of Prof. Guo Huadong was development of China’s Earth
Observation in the global context.
FIGURE 19. Prof. Michael Goodchild
presenting ideas on enhancing the value of
remote sensing.
TPC Cochairs Clive Fraser, Jeff
Walker, and Mark Williams have coordinated the work for the technical program of IGARSS’13. After the Plenary
talks Clive Fraser presented a few figures
concerning the Symposium.
SOCIAL PROGRAM
The Social Program of IGARSS 2013
provided lots of opportunities for discussions with colleagues and for having fun:
◗ Welcome Reception on Sunday
FIGURE 20. Dr. Rob Vertessy discussing
◗ Young Professionals Luncheon on
the role of remote sensing in monitoring
FIGURE 21. TPC Cochair Clive Fraser
Tuesday
severe weather in Australia.
providing information on IGARSS’13.
◗ Women in Geosciences and Remote
Sensing Reception on Tuesday
The Awards and Recognitions Ceremony was followed
◗ Melbourne Aquarium Reception on
after the tea break by presentations of the distinguished
Tuesday
Plenary speakers:
◗ Futsal Soccer Game on Wednesday
◗ Dr. Chris Pigram, Chief Executive Officer of Geoscience
◗ Technical Committee and Chapter Chairs Dinner on
Australia
Wednesday
◗ Prof. Guo Huadong, Director-General of the Institute
◗ Symposium Awards Banquet on Thursday.
of Remote Sensing and Digital Earth (RADI), Chinese
Academy of Sciences
YOU ARE INVITED TO ATTEND IGARSS 2014!
◗ Dr. Michael F. Goodchild, Emeritus Professor of
IGARSS 2014 will be held in Québec City, Canada, July
Geography at the University of California Santa Bar13–18, 2014, with Dr. Monique Bernier as the General Chair.
bara, USA
The symposium will be held in conjunction with the 35th
Canadian Symposium on Remote Sensing and the theme
◗ Dr. Rob Vertessy, Director of Meteorology and Chief
is “Energy and Our Changing Planet”. Please visit http://
Executive Officer of the Australian Government Bureau
____
igarss2014.com/Welcome.asp for further information.
of Meteorology.
58
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Multi-Temp 2013
Held June 25–27 in Canada
I
t has been 40 years since the launch of ERTS-1 (Landsat 1). At first, remote sensing research focused on
improving individual image quality. Gradually, it became
apparent that comparing images over time might yield
information necessary to manage environments at many
scales and over varying time periods. Especially, the years
since 2000 have seen ever-increasing imagery available at
different resolutions, and different time steps, from different agencies. Ever-improving techniques for radiometric equivalency have allowed more reliable discernment
of environmental change. As images are shown to hold
the keys to answering more and more questions, increasing numbers of researchers are becoming involved in the
emerging multi-temporal image analysis field. And of
course the need to manage change in our world is spurring on all of these developments.
The “Multi-Temp” workshop series has been in the
forefront of these developments. The first Multi-Temp was
held in 2001 in Trento, Italy; workshops have been held
every two years since, alternating between a European and
North American destination. Multi-Temps have been held
in Italy, Belgium, and the USA (Connecticut and Mississippi). Lorenzo Bruzzone (Trento, Italy), Pol Coppin (Leuven, Belgium), Roger King (Mississippi State University,
USA) and Ross Lunetta (USA) have been instrumental in
keeping the workshops of high quality and growing, as the
Multi-Temp Permanent Steering Committee.
The most recent Multi-Temp, The 7th International
Workshop on the Analysis of MultiTemporal Remote Sensing
Images (MultiTemp 2013) was held on June 25–27, 2013,
at Banff Alberta in the Canadian Rockies, marking the
first occasion MultiTemp has been held in Canada. With
Mryka Hall-Beyer (University of Calgary, Canada) as General Chair, and Greg McDermid (University of Calgary)
heading the scientific committee, MultiTemp 2013 featured sixty-six oral presentations, fifty-two posters, three
plenary panels with invited speakers, as well as keynote
addresses. This edition of MultiTemp also benefitted
greatly from a partnership between the IEEE GRSS and
The Canadian Remote Sensing Society (CRSS), and its
strong representation on the local organizing committee
in the persons of Derek Peddle (University of Lethbridge,
Canada), Joe Piwowar (University of Regina, Canada) and
SEPTEMBER 2013
Greg McDermid (University of Calgary). The committee
was rounded out by Guillermo Castilla (University of Calgary), Julia Linke (University of Toronto), Jennifer Hird
(University of Calgary) with strong background support
from John Yackel (University of Calgary).
The timing of Multi-Temp 2013 coincided with the
successful deployment of Landsat 8, providing an opportunity to look back on the Landsat series and explore
improved future opportunities, and to find out about
the specific program features that will make using both
Landsat 8 images and the Landsat archive more productive for multi-temporal applications. To this end, MultiTemp 2013’s keynote address was delivered by Dr. Curtis
Woodcock of Boston University who has been involved
with Landsat over many years. The conference was
framed by the last panel, on Landsat data continuity and
science. This panel, organized by Warren Cohen (Oregon
State University) brought together the perspectives of Dr.
Woodcock and David Roy (South Dakota State University), Joanne White (Canadian Forest Service), Richard
Allen (University of Idaho) and Todd Schroeder (U.S.
Department of Agriculture Forest Service).
Multi-Temp 2013 featured a panel session each day, to
bring in wide expertise. In addition to the Landsat keynote and panel, the conference featured a panel of Editors-in-Chief of major remote sensing publications. This
allowed free interchange on issues from peer-reviewing,
publication quality, guidelines for new researchers to
emerging issues of electronic publication. Organized by
Derek Peddle (University of Lethbridge, Canada), Editors-in-Chief on the panel were Marvin Bauer (Remote
Sensing of Environment), Russ Congalton (Photogrammetric Engineering and Remote Sensing), Derek Lichti (ISPRS
Journal of Photogrammetry and Remote Sensing), Nicholas
Coops (Canadian Journal of Remote Sensing) and Bill Emery
(IEEE Geoscience and Remote Sensing Journals). EiCs Emery
and Congalton were unable to attend, so their presentations were given by Lorenzo Bruzzone and Derek Peddle,
respectively.
The final panel reflected the conference theme of “the
changing environment” and focused on the uses of multitemporal images in monitoring environmental effects
of the oil sands development in Alberta. Organized by
Multi-Temp 2013 Science Committee Chair Greg McDermid (University of Calgary), this panel included Kirk
Andries (Alberta Biodiversity Monitoring Institute),
59
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Multi-Temp Delegates, Banff Alberta Canada—June 2013.
Paola DeRose (Canada Centre for Remote Sensing), Jennifer Grant (Pembina Institute, Canada) and Shane Patterson (Alberta Environment and Sustainable Resource
Development).
The very broad applications of multi-temporal images
were reflected in the topic groupings at Multi-Temp 2013.
Each day saw a plenary session allowing all attendees to
hear about important new research in broad topic areas.
In addition, individual sessions were devoted to a variety
of specific topics. These ranged from “old favorites” like
image calibration/data processing, change detection and
vegetation dynamics/phenology through water/hydrology applications, forestry, environmental monitoring,
and agriculture/grasslands (all reflecting strong Canadian
interests). The final group was devoted to topics promising important future research, and becoming prominent because of the increasing availability of time series
of high resolution and radar images. These topics were
urban applications and SAR-microwave. Finally, a session
grouped particularly innovative applications including
data integration and representations of time series.
A certificate was awarded for the best student oral
presentation, and went to Brent Smith, of the University
of Calgary, for his paper entitled “Examination of firerelated plant succession within the Dry Mixedgrass subregion of Alberta using MODIS and Landsat imagery”
(coauthor Greg McDermid).
Multi-Temp 2013 had to overcome some drastic environmental change to happen at all. When choosing the
theme “Our dynamic environment” the committee had
no inkling of just how dynamic the local environment
would turn out to be, and the difficulties this would
cause! Four days before the workshop was scheduled to
open, southwestern Alberta experienced an extremely
60
heavy rainfall event which resulted in flooding in Calgary (fortunately, away from the airport) and closure of
the Trans-Canada Highway, the direct route from Calgary
to Banff. Faced with the possibility of having to move
or even cancel the conference, the committee was much
cheered by an e-mail from attendee Daniel Gann (Florida
International University) reporting that he had arrived in
Banff at the height of the emergency, albeit by an alternate route. The local committee felt almost like Noah
receiving the dove’s olive branch indicating it was all right
to come out of the Ark, the flood was over! The committee started calling Daniel “our man in Banff”. Thanks to
the tireless efforts of emergency workers, a regular shuttle
bus service was running from the airport to Banff by the
day before the conference, and regular highway traffic
was restored before the conference ended. Despite these
challenges, almost all registered delegates arrived and
resulted in a full conference as planned.
On the bright side, the road closure meant that MultiTemp 2013 delegates saw Banff at its best, with brilliant
sunshine, quiet streets, uncrowded sights and un-rushed
services. Thanks to everyone who overcame difficulties,
and to all the people at the Banff Centre who made sure
we were welcome and superbly cared for. We look forward
to the next installment of Multi-Temp!
A complete program with abstracts for Multi-Temp
2013 can be downloaded from http://geog.ucalgary.ca/
multitemp2013/.
___________ Multi-Temp 2013 presenters and others working in this field will have the opportunity to
submit articles to special issues of IEEE J-STARS and the
Canadian Journal of Remote Sensing. The Multi-Temp 2013
Proceedings volume will be available on the IEEE Xplore
later this year.
GRS
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
GRSS MEMBER HIGHLIGHTS
GRSS Past President Jon Atli Benediktsson
Recognized As Icelandic IEEE/VFI
Electrical Engineer of the Year
P
rof. Jon Atli Benediktsson, IEEE GRSS Past President
was awarded on May 29 the Joint Icelandic Society
of Chartered Engineers (VFI) IEEE and 2013 Electrical
Engineer of the Year Award. The joint award was established in 2005 and is presented biannually. It is granted
to recognize excellent performance in the field of electrical engineering or as a recognition to electrical engineers who have made significant contributions to Icelandic society.
Prof. Benediktsson received the award “for outstanding technical achievements and world-wide recognition in the field of remote sensing, and in research
and manufacturing of medical equipment that measures—without invasion—oxygen saturation in the
veins of the eye.”
FIGURE 1. From the Award Ceremony (from left): Prof. Jon
Atli Benediktsson, IEEE/VFI Electrical Engineer of the Year, Mr.
Saemundur Thorsteinsson, IEEE Iceland Section Chairman and
Prof. Martin Bastiaans, IEEE Region 8 Chairman.
A National Society Agreement between IEEE and
VFI was signed in 2002 but IEEE has made such agreements with National Engineering Societies all over the world.
PROF. BENEDIKTSSON
The purpose of the agreements
RECEIVED THE AWARD FOR
is to encourage cooperation,
OUTSTANDING TECHNICAL
and coordinated joint activities
ACHIEVEMENTS IN REMOTE
between the country’s National
Society, and the local countries’
SENSING AND FOR MEDICAL
IEEE section. One of the coopEQUIPMENT THAT MEAerative activities has been the
SURES OXYGEN SATURATION
establishment of joint awards
IN THE EYE.
between the National Society
and IEEE.
FIGURE 2. From the Award Ceremony: Prof. Jon Atli Benediktsson, IEEE/VFI Electrical Engineer of the Year.
SEPTEMBER 2013
61
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
GRSS Members Elevated to the Grade of
Senior Member in the Period March–June 2013
MARCH:
APRIL:
JUNE:
Mudrik Alaydrus
Indonesia Section
Raja Boddu
Hyderabad Section
Bruce Campbell
Washington Section
Ulf Hanebutte
Seattle Section
Nettie La Belle-Hamer
Alaska Section
Fabio Pacifici
High Plains Section
Yuqi Bai
Beijing Section
Francesca Bovolo
Italy Section
Alejandro Frery
Bahia Section
Philip Hall
Victorian Section
Hrishikesh Kulkarni
Bombay Section
Maneesha Ramesh
Kerala Section
Zvonimir Sipus
Croatia Section
Ahmad Abawi
San Diego Section
Naif Alajlan
Saudi Arabia Section
Gilbert Carmona
Metropolitan Los Angeles
Section
Christian Debes
Germany Section
Roland Romeiser
Miami Section
Rama Nidamanuri
Kerala Section
Salvatore Stramondo
Italy Section
Senior membership has the following distinct benefits:
◗ The professional recognition of your peers for technical
and professional excellence.
◗ An attractive fine wood and bronze engraved Senior
Member plaque to proudly display.
◗ Up to $25.00 gift certificate toward one new Society
membership.
◗ A letter of commendation to your employer on the
achievement of Senior Member grade (upon the request
of the newly elected Senior Member).
◗ Announcement of elevation in Section/Society and/or
local newsletters, newspapers and notices.
◗ Eligibility to hold executive IEEE volunteer positions.
◗ Can serve as Reference for Senior Member applicants.
◗ Invited to be on the panel to review Senior Member
applications.
◗ Eligible for election to be an IEEE Fellow.
Applications for senior membership can be obtained
from IEEE website: _______________________
https://www.ieee.org/membership_
services/membership/senior/application/index.html.
__________________________________ You
can also visit the GRSS website: http://www.grss-ieee.org.
62
GRS
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
CALENDAR
See also HTTP://WWW.IEEE.ORG/CONFERENCES_EVENTS/INDEX.HTML
HTTP://WWW.TECHEXPO.COM/EVENTS
___________________________________________ or ___________________________
2013
2014
SEPTEMBER
DECEMBER
MAY
THE ASIA-PACIFIC CONFERENCE
ON SYNTHETIC APERTURE
RADAR (APSAR 2013)
September 23–27, 2013
Tsukuba, Japan
http://www.apsar2013.org/
9TH INTERNATIONAL
CONFERENCE ON MICROWAVES,
ANTENNA, PROPAGATION
& REMOTE SENSING
(ICMARS 2013)
December 11–14, 2013
Jodhpur, India
http://www.icmars2013.org
IEEE RADAR CONFERENCE:
FROM SENSING
TO INFORMATION
May 19–23, 2014
Cincinnati, Ohio, USA
http://www.radarcon2014.org
OCTOBER
URSI COMMISSION F,
MICROWAVE SIGNATURES 2013
SPECIALIST SYMPOSIUM ON
MICROWAVE REMOTE SENSING
OF THE EARTH, OCEANS, AND
ATMOSPHERE (URSI-F 2013)
October 28–31, 2013
Espoo (Helsinki), Finland
http://frs2013.ursi.fi/
GRS
Call for Papers
2014 IEEE Radar Conference:
From Sensing to Information
19-23 May 2014
Cincinnati, Ohio (USA)
Cincinnati Marriott at RiverCenter
10 th European Conference on
Synthetic Aperture Radar
03-05 June 2014 - Berlin, Germany
Tutorials: 02 June 2014
General Chair:
Prof. Brian Rigling – Wright State University
Technical Chair:
Dr. Muralidhar Rangaswamy – US Air Force Research Lab
GRSS Liaison:
Prof. Joel Johnson – The Ohio State University
Abstract submission: 18 October 2013
(Up to 4 pages with figures)
Author notification: 20 January 2014
Final papers: 21 February 2014
(Up to 6 pages with figures)
EUSAR is Europe's leading forum dedicated to SAR
techniques, technology and applications related technologies
with an international audience. We invite you to participate
in this world-class scientific event by submitting a paper.
This will be a unique opportunity for you to present your
research results, innovations and technologies to the world.
Draft Paper Submission Deadline:
October 31, 2014
Call for Exhibition and Sponsoring:
Please refer to www.eusar.de for details.
EUSAR 2014 General Chair: Manfred Zink, DLR
EUSAR 2014 Technical Chair: Gerhard Krieger, DLR
Web Address:
http://www.radarcon2014.org
SEPTEMBER 2013
63
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
AD INDEX
The Advertisers Index contained in this issue is compiled as a service to our readers and advertisers:
the publisher is not liable for errors or omissions although every effort is made to ensure its accuracy.
Be sure to let our advertisers know you found them through IEEE Geoscience and Remote Sensing Magazine.
IEEE Marketing Department
www.ieee.org/digitalsubscriptions
CVR 4
James A. Vick
Sr. Director,
Advertising
+1 212 419 7767;
Fax: +1 212 419 7589
[email protected]
_____________
Marion Delaney
Advertising Sales
Director
+1 415 863 4717
Fax + 1 415 863 4717
[email protected]
______________
Susan E. Schneiderman
Business Development
Manager
+1 732 562 3946;
Fax: +1 732 981 1855
[email protected]
_____________
PRODUCT
ADVERTISING
MidAtlantic
Lisa Rinaldo
Phone: +1 732 772 0160
Fax: +1 732 772 0164
[email protected]
_____________
NY, NJ, PA, DE, MD, DC,
KY, WV
New England/
South Central/
Eastern Canada
Jody Estabrook
Phone: +1 774 283 4528
Fax: +1 774 283 4527
[email protected]
_____________
ME, VT, NH, MA, RI, CT,
AR, LA, OK, TX
Canada: Quebec,
Nova Scotia,
Newfoundland,
Prince Edward Island,
New Brunswick
Southeast
Thomas Flynn
Phone: +1 770 645 2944
Fax: +1 770 993 4423
[email protected]
_____________
VA, NC, SC, GA, FL, AL,
MS, TN
Midwest/Central Canada
Dave Jones
Phone: +1 708 442 5633
Fax: +1 708 442 7620
[email protected]
_____________
IL, IA, KS, MN, MO, NE,
ND, SD, WI, OH
Canada: Manitoba,
Saskatchewan, Alberta
Midwest/Ontario,
Canada
Will Hamilton
Phone: +1 269 381 2156
Fax: +1 269 381 2556
[email protected]
______________
IN, MI. Canada: Ontario
West Coast/
Mountain States/
Western Canada
Marshall Rubin
Phone: +1 818 888 2407
Fax: +1 818 888 4907
[email protected]
_____________
AZ, CO, HI, NM, NV,
UT, AK, ID, MT, WY, OR,
WA, CA
Canada: British Columbia
Europe/Africa/
Middle East/Asia/Far
East/Pacific Rim
Heleen Vodegel
Phone:
+44 1875 825 700
Fax: +44 1875 825 701
[email protected]
_____________
Europe, Africa, Middle
East, Asia, Far East,
Pacific Rim, Australia,
New Zealand
RECRUITMENT
ADVERTISING
MidAtlantic
Lisa Rinaldo
Phone: +1 732 772 0160
Fax: +1 732 772 0164
[email protected]
_____________
NY, NJ, CT, PA, DE, MD,
DC, KY, WV
New England/
Eastern Canada
Liza Reich
Phone: +1 212 419 7578
Fax: +1 212 419 7589
[email protected]
__________
ME, VT, NH, MA, RI
Canada: Quebec,
Nova Scotia,
Newfoundland,
Prince Edward Island,
New Brunswick
Southeast
Cathy Flynn
Phone: +1 770 645 2944
Fax: +1 770 993 4423
[email protected]
_____________
VA, NC, SC, GA, FL, AL,
MS, TN
Midwest/South Central/
Central Canada
Darcy Giovingo
Phone: +1 224 616 3034
Fax: +1 847 729 4269
[email protected]
_____________
AR, IL, IN, IA, KS, LA, MI,
MN, MO, NE, ND, SD,
OH, OK, TX, WI
Canada: Ontario,
Manitoba, Saskatchewan,
Alberta
West Coast/Southwest/
Mountain States/Asia
Tim Matteson
Phone: +1 310 836 4064
Fax: +1 310 836 4067
[email protected]
_____________
AZ, CO, HI, NV, NM, UT,
CA, AK, ID, MT, WY,
OR, WA
Canada: British Columbia
Europe/Africa/
Middle East
Heleen Vodegel
Phone:
+44 1875 825 700
Fax: +44 1875 825 701
[email protected]
_____________
Europe, Africa, Middle
East
64
SEPTEMBER 2013
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
ʹͲͳ͵

͵ͲǦͶʹͲͳ͵
ǡǡ
Ǧ
ʹͲͳ͵Ǥ
Ǧ Ǧ
ǡǤ
ȀǡǡǤ
ʹͷǤ

Ǥ ǡ
ǡ[email protected] ʹͺǡ
ʹͲͳ͵Ǥ ǡ Ǥ
͵ͳǡ ʹͲͳ͵Ǥ ǡ
ǡǣȀȀǤǦǤȀǤ
Geoscience and Remote Sensing South Italy Chapter
Geoscience and Remote Sensing Spain Chapter
M
q
M
q
M
q
M
q
MQmags
q
M
q
M
q
M
q
M
q
MQmags
q
Instant Access to IEEE Publications
Enhance your IEEE print subscription with
online access to the IEEE Xplore® digital library.
Q
Download papers the day they are published
Q
Discover related content in IEEE Xplore
Q
“IEEE is the umbrella that
allows us all to stay current
with technology trends.”
Dr. Mathukumalli Vidyasagar
Head, Bioengineering Dept.
University of Texas, Dallas
Signiﬁcant savings over print with an online
institutional subscription
Start today to maximize
your research potential.
Contact: ________________
[email protected]
www.ieee.org/digitalsubscriptions
M
q
M
q
M
q
M
q
MQmags
q