Dr. Chan,
Please pass on the following summary of responses to CVNet. Thank you!
On 14 Aug 1997, I submitted the following post to CVNet (c/o Hoover Chan)
entitled: "Q: Colored Spot Detectibility"
****************
I'd appreciate any reference pointers or immediate answers to the following.
An observer views a small spot, in both chromatic and lightness contrast,
against some "uniform" background, under photopic conditions. The spot is
distant (so depth cueing not a factor), and its diameter subtends anywhere
from 5-25 arcminutes.
My question is how to determine the joint color/lightness contrast
perceptibility threshold (Ct).
There's plenty of classical luminance-only experiments applying to just
such situations, which suggests a Weber Law form Ct of about 0.03
(near-foveal), for spots greater than 10 arcminutes (i.e., at the crossing
between constant-flux and constant-luminance size regimes). But what about
for color AND brightness?
In my ignorance, I'm tempted to employ a color difference metric, e.g., CMC
or CIE94, and scale it accordingly against an achromatic Ct of 0.03.
However, I wonder, at such spatial scales here, if color segregation is
occurring near contrast threshold. Is this approach appropriate? Is there
something better, most particularly relevant experimental data?
********************
<<<<<<------>>>>>>
I received the following helpful replies summarized below:
(Courtesy of Jim Thomas; Thomas@mail1.lifesci.ucla.edu):
=============
"We conducted a study on this problem for the US Coast Guard. Targets, of
the size range in which you are interested, differed from the background in
luminance, chrominance, or both. A model was developed, tested and
parameterized using CRT simulations, and validated in a field study using
actual targets viewed at distances. The results are contained in the
following technical reports:
L. A. Olzak, J. P. Thomas, & H. Stanislaw (1987). Development of a
Chromatic/Luminance Contrast Scale. United States Department of
Transportation, United States Coast Guard. November, 1987. (DOT Report
No. CG-D 12-87)
L. A. Olzak (1989). Development of a Chromatic/Luminance Contrast
Scale: Field Validation. United States Department of Transportation,
United States Coast Guard. (DOT Report No. CG-D 2-89)"
Frank comments: --> I retrieved these references. Excerpt from the
abstract: "The model quantitatively describes the interrelationships among
detectibility, target size, target luminance, target chromaticity,
background chromaticity, and background luminance". The data is
presented, and the model receives inputs w.r.t. untransformed XYZ space.
============
(Courtesy of Ewen King-Smith; king-smith.1@osu.edu):
============
"David Carden and I published some data (1976, J. Opt. Soc. Am., 66, 709-717)
which might have some relevance, measuring thresholds and spatial
integration for spectral lights on a white background. Results were
analyzed in terms of luminance and opponent-color contributions. Spatial
integration for the red-green color system was considerably greater (perhaps
a factor of 2 in critical diameter) than for the luminance system; spatial
integration for the blue-yellow system was even bigger (about 5 times that
for the luminance system). See Fig. 12 of that paper."
Frank comments explanatorily: --> Contrast thresholds as function of
wavelength and spot size, for differing flash periods. Higher thresholds
for smaller spots, with size of spatial integration field indicated
whereupon threshold transition to constant with increasing size.
============
(Courtesy of Rhea Eskew, ESKEW@neu.edu)
============
"I would strongly suggest you calculate the cone contrast vector of your
test spot {deltaL/L, deltaM/M, deltaS/S} first. If you're unsure of
how to do that I can make some suggestions (one of which would be to
see David Brainard's appendix in Human Color Vision by Kaiser & Boynton).
Once you know that, there are several possible next steps. One would be
to calcuate the projection of your test vector on theoretical color
rg, yb, and luminance mechanisms. The three mechanisms
weights I would use (these are from a review chapter of mine in press,
but its likely that most other published mechanisms would produce
similar results) are:
L M S
RG: -.70, .72, -.02
YB: -.55, -.25, 0.8
Lum: 0.78, .37, 0
(in other words, you take the dot product of your test cone contrast vector
with each of those three vectors).
The .03 you mention is a reasonable estimate of the luminance
mechanism's threshold here; rg's is about 3-fold lower (see our paper
in Nature, 1993, vol 361, 348-350 -- but note we use a somewhat different
definition of contrast, not one I would recommend to you).
[Full citation: A. Chaparro, C.F. Stromeyer III, E.P. Huang, R.E.
Kronauer, and R.T. Eskew, "Colour is what the eye sees best", Nature, Vol.
361 28 January 1993, pp. 348-350] The yellow-blue threshold is at least as
high as the luminance one, and probably several times higher.
Unless the luminance contrast is several times greater than the rg contrast,
therefore, its likely the spot is detected by the red/green mechanism, and
you need not worry about a joint metric.
In addition to the Nature paper, you might also look at Chaparro, A.,
Stromeyer, C.F. III, Kronauer, R.E., & Eskew, R.T. Jr. (1994) Separable
red-green and luminance detectors for small flashes. Vision Research,
34, 751-762. for some threshold data."
=================
(Courtesy of Chien-Chung Chen; chen@skivs.ski.org)
=================
"Actually, there are some studies have been done related in your
work. The most relevant work is
> Cole, G. R.; Stromeyer, C. F.; Kronauer, R. E.(1990).
Visual interactions with luminance and chromatic stimuli.
Journal of the Optical Society of America A, 1990 Jan, v7 (n1):128-140
About defining chromatic and luminance contrast,
Poirson & Wandell
> Poirson, Allen B.; Wandell, Brian A.(1993)
Appearance of colored patterns: Pattern-color separability.
Journal of the Optical Society of America A, 1993 Dec, v10
(n12):2458-2470.
and me
> Chen, C.-C. (1996)
Chromoluminance pattern vsion: Masking experiments and divisive
inhibition models.
University of California, Santa Barbara
> Chen, C.-C., Foley, J. M. & Brainard, D. H. (1997).
Detecting chromatic patterns and chromatic patterns pedestals.
IS&T Proceedings: Optics & Imaging in the Information age, 19-24.
have made an effort on that. And a more elaborated review was
provided by Brainard in the appendix of the book
> Kaiser, Peter K.; Boynton, Robert M (1996)
Human Color Vision. OSA Press.
The basic idea is that, based on the cone fundamental measurement
(e.g. Smith & Pokorny, 1975), you can computed how much excitation
is produced in each cone class by your stimulus. Comparing the cone
excitations for the background and the test. You are able to
calculate the cone contrasts in each cone type of the test against
the background. The cone contrasts are pooled together by a pooling
rule. In this way, you can calculate the contrast of your
stimuli regardless it is luminance contrast or chromatic contrast.
It is rather difficult to put equations in E-mail. But you can
find all the equations you want in the above references."
======================
(Courtesy of Bosco Tjan; tjan@mpik-tueb.mpg.de)
======================
"There is no general approach of do this because color and luminance are two
orthrogonal sensory inputs. They are essentially quantities in different units
(like trying to add 1 kg to 5 seconds).
The way of combining them depends on your application. For example, if your
application concerns with painting an aircraft, there will be several
parameters you can used to specify the final luminance and chromatic contrast
of a paint (against some background). You can then what these parameters do to
detectability.
If your concern is more fundamental (with respect to human perception), cone
contrast space may be something to use. (See Appendix IV of "Human Color
Vision", 2nd Edition by Kaiser and Boynton).
Finally, you can express the two (color and luminance contrast) in terms of the
multiples of their respective detection thresholds (e.g. "the stimulus is of
1.5 luminance threshold and .7 red/green threshold"). In these units, it has
been show that color and luminance signals added in a probability summation
fashion (independent coding), even for high-level task such as reading (Legge
GE, Parish DH, Luebker A & Wurn LH (1990) "Psychophysics of reading XI.
Comparing color contrast and luminance contrast." Journal of the Optical
Society of America, A7, 2002-2010)."
=======================
Thanks also to Peter D. Gowdy for his suggestion (not included above).
Additional references I've traced from others or otherwise uncovered:
D.B. Judd, "Color in Visual Signal.....[I've only the "figure 7" from a xerox]":
Apparently pertaining to the WUV color space. "The ability of observers
to detect red-green (U) differences went to zero for circular fields
subtending about 2 minutes of arc. ... Also the ability of observers to
detect violet-green yellow (V) differences went to zero at about 8 minutes
of arc."
A.L. Nagy and R.R. Sanchez, "Critical color differences determined with a
visual search task", J.Opt.Soc.Am.A, 7:7 (1990).
Abstract: "Response times were measured for a visual search task [colored
target and distractor dots on a dark background-Frank] in which the
observer was required to find a target that differed from distracting
stimuli only in color. In the first experiment the search time was
measured as a function of display density for both small and large color
differences. With small color differences response times increased with
display density, indicating a serial search, but with large color
differences response time was constant, indicating a parallel search. In
the second experiment the color difference required for parallel search was
measured in eight different directions from the distractor chromaticity.
These color differences were much larger than threshold color differences
and were not well represented by the ellipse [MacAdam's] used to describe
the threshold contour around a point in color space."
A.L. Nagy, R.T. Eskew, Jr., and R.M. Boynton, "Analysis of color-matching
ellipses in a cone-excitation space", J.Opt.Soc.Am.A, 4:4 (1987).
Abstract: "Color-discrimination ellipses derived from the variability of
color-matching data of six observers are analyzed in a normalized
constant-luminance cone-excitation space. The analysis shows that the
ellipses do not vary significantly in shape with chromaticity, observer, or
experimental conditions. The discrimination contours are predictable from
the thresholds on the two cardinal [L-2M, S] axes of this space; these are
used to normalize the data at each chromaticity for each observer.
Thresholds on these two axes vary with chromaticities, individuals, and
experimental conditions in accordance with simple and familiar laws."
++++++++++++
Regards,
Frank J. Iannarilli, Jr. franki@aerodyne.com
Aerodyne Research, Inc., 45 Manning Rd., Billerica, MA 01821 USA