AD-A250 705
absr-tr. , 5
AFOSR Grant No. AFOSR-89-0304
The Effects of Luminance Boundaries on Color Perception
Professor Richard E. Kronauer
Harvard University
Division of Applied Sciences
Pierce Hall 324
Cambridge, MA 02138
April 30, 1992
s
DTIC
ELECTE j
MAY 18 1992 I
Period Covered: March 15, 1991 to March 14, 1992
ANNUAL TECHNICAL REPORT
This document has been appioved
lot public release and sale; its
distribution is unlimited.
Dr. John F. Tagney
AFOSR/NL
Washington, DC 20332-6448
92 5 14 144 B
92“12987
UNCLASSIFIED
JN OF THIS PAGE
REPORT DOCUMENTATION PAGE
Form Approved
0MB No. 0704-0188
REPORT SECURITY CLASSIFICATION
Unclassified
2». SECURITY CLASSIFICATION AUTHORITY
2b. DECLASSIFICATION/CKIWNGRADING SCHEDULE
4. PERFORMING ORGANIZATION REPORT NUM8ER(S)
1b RESTRICTIVE MARKINGS
NONE
3 DISTRIBUTION/AVAILABILITY OF REPORT
Unlimited
S MONITORING ORGANIZATION REPORT NUMBER(S)
6a. NAME OF PERFORMING ORGANIZATION
Harvard University
6c ADDRESS {City, State, and ZIP Code)
Division of Applied Sciences
Cambridge, MA
8a. NAME OF FUNDING/SPONSORING
ORGANIZATION
AFOSR
6b. OFFICE SYMBOL
(If applicable)
8b. OFFICE SYMBOL
(If applicable)
8c ADDRESS (Oty, State, and ZIP Code)
AFOSR/NL
Washington, DC 20332-6448
11. TITLE (Include Security Clauilication)
7a. NAME OF MONITORING ORGANIZATION
Office of Naval Research o j; V
Resi^ient' Representative
7b. ADDRESS (C/fy, State, and ZIP Code)
Charles S. Draper Laboratories
555 Technology Square 'SrC-
Cambridge, MA 02139 _
9 PROCUREMENT INSTRUMENT IDENTIFICATION NUMBER
AFOSR-89-0304
10. SOURCE OF FUNDING NUMBERS
PROGRAM
ELEMENT NO.
PROJECT
NO.
The Effects of Luminance Boundaries on Color Perception (unclassified) _
12. PERSONAL AUTHOR(S)
Kronauer, R.E., Stromeyer, C.F. Ill, Chaparro, A., and Eskew, R.T., Jr._
13a. TYPE OF REPORT 13b. TIME COVERED 14 DATE OF REPORT (Year, Month, Day) 15. PAGE COUNT
Annual Technical Rpt from 3/15/91 TO 3/14/9 2 April 30, 1992 11
COSATI COOES
SUB-GROUP
18. SUBJECT TERMS (Continue on reverse if necessary and identify by block number)
19. ABSTRACT (Continue on reverse if necessary and identify by block number)
Extensive measurements were made for detecting luminance and red-green flashes in the
center of a bright yellow field. Thresholds, plotted in L and M-cone contrast coordinates,
indicate that chromatic flashes are more visible than luminance flashes even at very small
size (2' diameter). Over a wide range of flash diameters and durations the chromatic flashes
are detected with considerably higher efficiency (in units of cone contrast energy) than the
most detectable luminance stimuli (small drifting gratings). The higher gain of the chromatic
mechanisms has important physiological implications and is potentially useful in display
technology. Detailed studies with luminance and chromatic stimuli suggest that the chromatic
mechanisms have a constant spectral tuning, even for spots as small as 2': the chromatic
response is determined by a constant , equally weighted difference of L and M cone contrast.
A suprathreshold luminance flash (a pedestal) facilitates detection of a coincident
chromatic flash. Earlier studies suggested that the facilitation will grow strongly when
the stimuli were decreased in size. In contrast, we find that facilitation is constant (2-3x)
for stimuli from 2 ' to 2° diameter.
DO Form 1473, JUN 86
Previous editions are obsolete.
SECURITY CLASSIFICATION OF THIS PAGE
5 m
•AFOSR Grant No. 89-0304
Annual Technical Report 1991/92
1
The Effects of Luminance Boundaries on Color Perception
Period Covered: March 15, 1991 to March 14, 1992
This report covers our activities since March 15, 1991. Our
main accomplishments have been to: 1) Finish experiments on
detection of small spots by luminance and red-green mechanisms,
examining the role of a luminance pedestal in facilitating
chromatic detection at small spot sizes. 2) Show that the red-
green mechanism is more efficient than the luminance mechanism in
contrast detection. 3) Consider further the roles of cone-
selective adaptation and 'second-site' adaptation in controlling
the sensitivity of the red-green detection mechanism. 4) Continue
work on the spectral nature of the inputs to motion detection
mechanism.
On a large, bright yellow field we deliver a foveal flash
consisting of simultaneous incremental and decremental red and
green components. Detection thresholds are measured for many
different red:green amplitude ratios, and thresholds are plotted
■AFOSR Grant No. 89-0304
Annual Technical Report 1991/92
2
Using a 200 ms test flash, we demonstrated the existence of
the red-green mechanism by measuring detection contours for spots
over a size range of 2.3' to 1“^ diameter. At the largest size the
red-green mechanism is lOx more sensitive than the luminance
mechanism, and this ratio decreases as the spot is made smaller,
but the red-green mechanism is about 2x more sensitive even at the
smallest size.
The most efficiently detected stimulus on the bright yellow field
These experiments were extended to determine "What the eye
sees best?" measured in terms of contrast energy of the stimulus.
Watson, Barlow and Robson (1983) attempted to answer this for
stimuli in the luminance domain. They measured contrast energy
thresholds (the square of contrast integrated over the spatial and
temporal dimension of the test stimuli), using both incremental
spots and small patches of a drifting grating. The best grating
had an energy threshold 3x less than the best spot and identifies
approximately the shape of the most sensitive luminance receptive
field (one that matches the most efficient stimulus). We measured
spot luminance thresholds over a wide range of flash sizes and
durations. Our luminance spot energy thresholds agree very well
with those of Watson ai- Our best-detected red or green
chromatic spot has a cone contrast energy threshold 5-8x lower
than our best luminance spot and about 3-8x lower than Watson ai. '
ai.'s optimal grating. The higher sensitivity for the chromatic
stimulus is not solely due to better spatial and temporal
integration in the chromatic pathways, for there is a clear
AFOSR Grant No. 89-0304
Annual Technical Report 1991/92
3
chromatic advantage even when the chromatic spot is matched in
size and duration to the optimal luminance spot.
The higher sensitivity to color is surprising, but may be
consistent with recent findings of R.M. Shapley and colleagues
(largely unpublished) on retinal ganglion cells. It is well know
that the M ganglion cells have higher contrast gain than do the
color-opponent P cells when tested with a luminance grating.
However, when tested with a chromatic grating, matched in cone
contrast to the luminance grating, then the P and M cell have
similar contrast gains. The P cells are much more numerous, and
their receptive fields often have largely overlapping L and M
areas with similar weights but opposite signs, making them
especially sensitive to color, and much less sensitive to
luminance.
Our work is described in the enclosed paper (Chaparro ££. al.,
1992), which we intend to submit to Nature .
Chromatic facilitation by luminance pedestals at small spot size
A suprathreshold luminance flash of 1° diameter presented
simultaneously with a red or green equiluminant chromatic flash
facilitates the latter's threshold by 2x. Earlier work by Hilz,
Huppmann and Cavonius (1974) led us to suspect that the
facilitation would grow much larger when the pedestal and
chromatic flash were concomitantly reduced in size. However, we
find that facilitation remains approximately constant at 2-3x, as
measured by forced-choice methods, for flashes from 2.3' to 1°
diameter.
AFOSR Grant No. 89-0304
Annual Technical Report 1991/92
4
The 'variable tuning' hypothesis advanced by Finkelstein and
Hood (1984) postulates that the mechanism used to detect chromatic
spots will change in its spectral sensitivity with variation of
the test spot size. Our results would seem to be at variance with
this view. Even for the smallest spots our red-green detection
contours have a slope of +1.0, indicating that L and M cones
contribute with equal and opposite weights (over the measured size
range 2.3' to 1°). A luminance pedestal has a polar vector
direction of +45° in the L',M' coordinates, approximately parallel
to the red-green contour--thus stimulating the red-green mechanism
very little. A 2.3' diameter luminance pedestal flash of even 25x
threshold does not mask the chromatic flash, consistent with the
view that the chromatic contour does have a constant slope of
approximately +1.0—not of variable tuning.
A surprising feature, remarked upon by Hood and Finkelstein
(1984) is that small colored flashes with a strong luminance
component (near the luminance axis in the L',M' plane), and
presumably below the threshold of the red-green mechanism, appear
colored reddish or greenish when only very slightly
suprathreshold. We have confirmed this and shown that the results
are quantitatively consistent with the view that the luminance
pedestal starts to facilitate the red or green chromatic test when
the pedestal reaches its own threshold. Thus there is a
facilitating interaction between luminance and chromatic
mechanisms that can explain the appearance of color of flashes
that are near the luminance axis. We have shown that the threshold
for identifying the hue of the test (red vs green) is well
AFOSR Grant No. 89-0304
Annual Technical Report 1991/92
5
predicted from the curve relating luminance pedestal amplitude and
chromatic (facilitated) detection thresholds.
We are presently writing up the results; the experiments are
completed.
Mechanisms of adaptation in the red-green pathways
Previously we (Stromeyer, et al., 1985) observed that on
bright chromatic adapting fields, the sensitivity of the red-green
detection mechanism is controlled by two adaptation mechanisms.
First, the L and M cones differentially adapt (or the L and M
cone-selective pathways differentially adapt) following Weber's
Law, so that the red-green detection contour maintains a slope of
about +1.0 in the L',M' coordinates for different colored adapting
fields. Second, there is also second-site adaptation, which
reduces sensitivity at an opponent site via response saturation
when the adapting field is strongly chromatic (producing an
extreme in the ratio L/M for mean adaptation). The second-site
adaptation cause the red-green detection contour to move outward
from the origin in the L',M' coordinates, reflecting reduced
sensitivity.
Kraus)copf and Gegenfurtner (1991) measured red-green
detection thresholds on different colored fields with test stimuli
restricted the equiluminant plane . They found surprisingly
little influence of field color on threshold and argued that
first-site adaptation plays little role in determining
equiluminant red-green sensitivity. We have shown, however,
(Es)cew, Stromeyer and Kronauer, 1992) that these results are
consistent with both first- and second-site adaptation. The
•AFOSR Grant No. 89-0304
Annual Technical Report 1991/92
6
difficulty associated with stimuli restricted to the equiluminant
plane has to do with the fact that, although this plane is 2
dimensional, one dimension is assigned to S cone stimulation (the
so-called blue-yellow axis) so that only one dimension remains for
the L/M cone system. Therefore only a highly limited view of L/M
cone interaction is seen in the isoluminant plane. Rather one
should measure the full red-green detection contour in the L',M'
cone contrast coordinates, since the effects of adaptation are
particularly evident near the unique L' or M' axis. When we
restrict our detection data to the isoluminant conditions of
Krauskopf and Gegenfurtner we find results very similar to theirs.
Since Krauskopf and Gegenfurtner's data were obtained at lower
adapting levels than we used, we intend this summer to make
thorough measurements of the red-green contour, in the L',M'
coordinates, for a range of adapting colors and mean levels, to
assess the role of the two postulated adaptation mechanisms.
Cone inputs for motion detection
A large part of our recent effort has been devoted to
carefully assessing the spectral nature of mechanisms detecting
motion. Thus far we have been examining the L and M cone signals.
The observer views a 1 cpd vertical red-plus-green grating that
moves left or right on a 3500 td yellow field. We typically
measure contrast thresholds for discriminating motion (left vs
right). As described in our previous progress report, we have
identified two spectral motion mechanisms, which are distinct from
the red-green hue mechanism. The latter hue mechanism has a
constant slope of +1.0 in the L',M' cone contrast coordinates
•AFOSR Grant No. 89-0304
Annual Technical Report 1991/92
7
(measured by simple detection or hue identification). The two
motion mechanisms are a luminance type mechanism and a spectrally-
opponent mechanism. The latter responds to the differences of L
and M cone contrast, but the M cone contribution drops rapidly as
velocity is increased. The luminance mechanism responds to the sum
of L and M cone contrast signals: at high velocities the two cone
types contribute with similar weights, while at low velocities the
mechanism becomes L-cone dominated. Similar temporally-dependent
properties of L and M inputs have been observed in macaque phasic
ganglion cells by Lee, Martin and Valberg (1989).
We have shown that the outputs of the luminance and
spectrally-opponent motion mechanisms eventually summate within
opponent motion mechanisms that are sensitive to the differences
of right and left motion components.
We have obtained very extensive measurements on three
observers, but still hsvs considersble data to collect, “specially
pertaining to the summation of signals from the different spectral
mechanisms.
References:
Chaparro, A., Stromeyer, C.F. Ill, Huang, E.P., Kronauer, R.E.,
and Eskew, R.T., Jr. (1992). What the eye sees best: Colour.
To be submitted to Nature (enclosed).
Finkelstein, M.A. and Hood, D.C. (1984). Detection and
discrimination of small, brief lights: variable tuning of
opponent channels. Vision Res . 24 , 175-181.
AFOSR Grant No. 89-0304
Annual Technical Report 1991/92
8
Hilz, R.L., Huppma m, G. and Cavonius, C.R. (1974). Influence of
luminance contrast on hue discrimination. Journal of the
Optical Society of America ^ M/ 763-766.
Krauskopf, J. and Gegenfurtner, K. (1991). Adaptation and color
discrimination. In Valberg, A. and Lee, B.B. From Pigments to
Perception . N.Y., Plenum.
Lee, A.B., Martin, P.R. and Valberg, A. (1989). Sensitivity of
macaque retinal gar.glion cells to chromatic and luminance
flicker. J. Phvsiol . 414 . 223-243.
Stromeyer, C.F. Ill, Cole, G.R., and Kronauer, R.E. (1985).
Second-site adaptation in the red-green mechanism. Vision
Eai. 2^, 219-327.
Watson, A.B., Barlow, H.B. and Robson, J.G. (1983). What does the
eye see best? Nature 302 , 419-422.
Participating Professionals
Professor Richard E. Kronauer
Dr. Charles F. Stromeyer III
Dr. Rhea T. Eskew, Jr.
Dr. Alex Chaparro
Euill icatlPns and Publications in Progress
Chaparro, A., Stromeyer, C.F. Ill, Huang, E.P., Kronauer, R.E.,
and Eskew, R.T., Jr. (1992). What the eye sees "best":
Colour. To be submitted shortly to Nature (manuscript
enclosed).
. AFOSR Grant No. 89-0304
Annual Technical Report :991/92
9
Eskew, R.T., Jr., Stromeyer, C.F. Ill, and Kronauer, R.E. (1992).
The co.istancy of equiluminant red-green thresholds examined
j 1 two color spaces. In Proceedings of QSA Meeting : Advances
in Color Vision. Technical Digest (Optical Society of
America) 4., 195-197 (enclosed) .
Eskew, R.T., Jr., Stromeyer, C.F. Ill and Kronauer, R.E. (1992a).
The time course of the facilitation of chromatic detection by
luminance contours. Paper in progress (manuscript available
upon request).
Eskew, R.T., Jr., Stromeyer, C.F. Ill ana Kronauer, R.E. (1992b).
On the temporal chromatic impulse response function. Paper in
progress (manuscript available upon request).
Stromeyer, C.F. Ill, Lee, J. and Eskew, R.T., Jir. (1992).
Peripheral chromatic sensitivity for flashes: a post-
receptoral red-green asymmetry. Vision Research (In press)
(manuscript enclosed).
Papfixs bei ng delivered at recent professional meetings
(Abstracts enclosed).
Chaparro, A., Stromeyer, C.F. Ill, Kronauer, R.E., and Eskew,
R.T., Jr. (1992). The detection efficiency of chromatic
stimuli. Annual meeting of Association for Research in Vision
and Ophthalmology. Invest. Qphthal. & Visual Sci. 33/4 . 755.
Eskew, R.T., Jr., Chaparro, A., Stromeyer, C.F. Ill and Kronauer,
R.E. (1992). Facilitation of red-green detection by
luminance pedestals at small spot size. Annual Meeting of
Association for Research in Vision and Ophthalmology. Invest .
Qphthal. _ & Visual Science 33/4 , 702.
AFOSR Grant No. 89-0304
Annual Technical Report 1991/92
10
Eskew, R.T., Jr., Stromeyer, C.F. Ill, Chaparro, A., and Kronauer,
R.E. (1992). Why is chromatic sensitivity greater than
luminance sensitivity? Annual Meeting, Optical Society of
America, Albuquerque, NM.
Stromeyer, C.F. Ill, Kronauer, R.E., Eskew, R.T., Jr. (1992).
Relative temporal phase of L vs M cone signals within the
luminance motion mechanism. Annual Meeting of Research in
Vision and Ophthalmology, Sarasota, FL. Invest. Ophthal. &
Visual Sci . . 33/4 , 756.
