poster2000 - Center for Neural Science

Comments

Transcription

poster2000 - Center for Neural Science
#1255 SURFACE COLOR AND SPECULARITY: TESTING THE D’ZMURA-LENNIE-LEE MODEL
J. N. YANG & L. T. MALONEY, Department of Psychology and Center for Neural Science, New York University
3. PERTURBATION METHOD
Cue Promotion
-- von Helmholtz
There are many possible cues to the illuminant (Maloney, 1999), not all of which are
present in every scene. We treat illuminant estimation as a cue combination problem
and seek to determine which cues to the illuminant are used in particular scenes.
Last year (Yang, Maloney & Landy, 1999) we reported that information about the
illuminant conveyed by surface specularity influenced judgments of color appearance.
Base D65
Illuminant A
v’
Full Surface
Specularity
Dynamic Re-Weighting
Perturbed
Illuminant D65 (matte)
Illuminant A (specular)
ILLUMINANT CUE COMBINATION
base D65
target A
perturbed
u’
The first and third scene above are a single-matte scene illuminated under two different
illuminants. The middle scene is perturbed: all illuminant cues except specularity signal
D65 (the base illuminant) while all specularity cues signal A.
1. SPECULAR CUES
SPECULAR HHIGHLIGHT
IGHLIGHT
SPECULAR
CUE
The influence of the specularity cues can be quantified
as the ratio of the length of the solid vector (the effect
of perturbation) to the length of the dotted line connecting
the base and target conditions (the effect of changing the
illuminant):
We measure the observer’s achromatic setting in all three scenes.
There are currently two kinds of specularity-based algorithms for
estimating illuminant chromaticity.
If the observer’s achromatic setting for the perturbed scene is identical to that for the
base D65 scene, then the perturbation had no effect. The observer is not influenced by
the specular information.
In the first method, we use the chromaticity of isolated specular
highlights as an estimate of illuminant chromaticity.
If the observer’s achromatic setting for the perturbed scene is identical to that for the
target A scene, then only specularity influences the observer’s judgment. Perturbing
specularity is equivalent to changing the illuminant on the scene.
This specular highlight cue is available if a visual system can
identify neutral specular highlights in scenes.
The illuminant chromaticity estimate based on this specular highlight cue
can be contaminated by the color of the matte(non-specular) component
of a surface.
Specularity cues perturbed ...
I=
||
||
||
-
||
We expect that the achromatic setting for the perturbed scene will fall somewhere
between the achromatic settings for the base scene D65 and the achromatic setting
for the target scene A, and we use this to quantify the influence of the cue.
Single-Matte Scene
Illuminant D65
Single-Matte
out by forming a judgment about
the colors of bodies, eliminating
the differences of illumination by
which a body is revealed to us.
Specular
Highlights
Scene
Target A
Illuminant A
Multi-Matte
The algorithms differ mainly in the physical cues to the illuminant they employ.
Illuminant
Estimate
Uniform
Background
2. correct surface colors for the estimated illuminant chromaticity.
[I]n our observations with the
sense of vision, we always start
Our rendered scenes contain many potential illuminant cues,all signaling exactly
the same information about the illuminant.
In order to determine the influence of cues based on specularity, we need to perturb
the specularity cues so that they signal slightly discrepant information concerning the
illuminant.
Single-Matte
1. estimate the chromaticity of the illuminant,
4. EXPERIMENTAL CONDITIONS
JNY
Illuminant D65
Multi-Matte
Many computational models of surface color perception share a common structure:
The roles of the two illuminants can be reversed with A as base, D65 as target.
D’ZMURA-LENNIE-LEE CUE
5. RESULTS
Lee (1986) and D’Zmura & Lennie (1986) independently
proposed methods for removing the ‘matte’ contamination.
6. CONCLUSIONS
Both methods require that there be two or more surfaces with
distinct matte components with some specularity in the scene.
The scene to the right satisfies this condition. The scene above it
does not. The apples all share the same matte component.
0.52
v’
v’
We compare surface color perception in scenes where specular
objects have a single common matte component (Single-Matte Scenes)
and where they have multiple distinct matte components
(Multi-Matte Scenes).
Multi-Matte Scene
D65
D65
A
0.16
Stimulus Characteristics:
Observers viewed simulated (rendered) binocular
scenes comprising a flat background and 11 spheres.
All surfaces were Matte-Specular (Shafer, 1985) with matte
component matched to specific chips taken from the
Nickerson-Munsell collection.
In the Single-Matte Scenes, all sphere surfaces shared a single
matte component, in Multi-Matte Scenes they had 11 distinct matte
components.
Scenes were rendered under either of two reference illuminants,
A and D65 ( Wyszecki & Stiles, 1982).
Task: achromatic matching.
A
Single-Matte
0.24
0.16
u’
0.16
0.24
0.52
GT
CHF
D65
D65
0.52
0.24
0.16
u’
0.24
JA
EC
JA
EC
Landy, M. S., Maloney, L. T., Johnston, E. J. & Young,
M. (1995), Measurement and modeling of depth cue
combination: In defense of weak fusion. Vision
Research, 35, 389-412.
A
GT
CHF
0.52
v’
v’
0.16
u’
0.24
0.16
u’
0.24
Lee, H.-C. (1986), Method for computing the scene
illuminant chromaticity from specular highlights.
JOSA A, 3, 1694-1699.
Maloney, L. T. (1999), Physics-based models of surface
color perception. In Gegenfurtner, K. R. & Sharpe,
L. T. [Eds] (1999), Color Vision: From Genes to
Perception. Cambridge, UK: Cambridge University
Press, 387-418.
0.45
0.45
REFERENCES
D’Zmura, M. & Lennie, P. (1986), Mechanisms of color
constancy. JOSA A, 3, 1662-1672.
0.45
0.45
A
u’
v’
v’
Multi-Matte
Apparatus: Observers viewed stimuli in a
computer-controlled Wheatstone stereoscope.
u’
0.52
D65
We conclude that the visual system is not making use of the D’Zmura-Lennie-Lee
specular cue in these scenes.
0.45
0.45
The D’Zmura-Lennie-Lee specularity cue is available in the Multi-Matte scenes but is
weak or absent in the Single-Matte scenes.
Specularity had no influence on achromatic performance in the Multi-Matte scenes.
v’
v’
2. EXPERIMENTAL DESIGN
GT
CHF
0.52
BRM
EC
0.52
Surface color appearance is affected by the chromaticity of the specular component
of surfaces in some scenes, under some illuminants (Yang, Maloney & Landy, 1999).
We measured achromatic matching performance in two classes of scenes containing
evident specular cues to the illuminant: Single-Matte and Multi-Matte.
0.45
0.45
A
BRM
EC
0.52
GT
CHF
0.16
u’
0.24
0.16
u’
0.24
Maloney, L. T. & Yang, J. N. (in press), The illumination
estimation hypothesis. In Mausfeld, R. & Heyer, D.
[Eds] (in press) Colour Vision: From Light to Object.
Oxford: Oxford University Press.
Yang, J. N., Maloney, L. T. & Landy, M. S. (1999),
Analysis of illuminant cues in simulated scenes
viewed binocularly. IOVS, 40.

Similar documents