Washington University in St. Louis Washington University in St. Louis

Washington University Open Scholarship Washington University Open Scholarship

All Computer Science and Engineering Research Computer Science and Engineering

Report Number: WUCSE-2007-18

2007

The Effect of Object Color on Depth Ordering The Effect of Object Color on Depth Ordering

Reynold Bailey, Cindy Grimm, Christopher Davoli, and Richard Abrams

The relationship between color and perceived depth for realistic, colored objects with varying

shading was explored. Background: Studies have shown that warm-colored stimuli tend to

appear nearer in depth than cool-colored stimuli. The majority of these studies asked human

observers to view physically equidistant, colored stimuli and compare them for relative depth.

However, in most cases, the stimuli presented were rather simple: straight colored lines, uniform

color patches, point light sources, or symmetrical objects with uniform shading. Additionally, the

colors were typically highly saturated. Although such stimuli are useful for isolating and

studying depth cues in certain contexts, they... Read complete abstract on page 2. Read complete abstract on page 2.

Follow this and additional works at: https://openscholarship.wustl.edu/cse_research

Part of the Computer Engineering Commons, and the Computer Sciences Commons

Recommended Citation Recommended Citation Bailey, Reynold; Grimm, Cindy; Davoli, Christopher; and Abrams, Richard, "The Effect of Object Color on Depth Ordering" Report Number: WUCSE-2007-18 (2007). All Computer Science and Engineering Research. https://openscholarship.wustl.edu/cse_research/122

Department of Computer Science & Engineering - Washington University in St. Louis Campus Box 1045 - St. Louis, MO - 63130 - ph: (314) 935-6160.

This technical report is available at Washington University Open Scholarship: https://openscholarship.wustl.edu/cse_research/122

The Effect of Object Color on Depth Ordering The Effect of Object Color on Depth Ordering

Reynold Bailey, Cindy Grimm, Christopher Davoli, and Richard Abrams

Complete Abstract: Complete Abstract:

The relationship between color and perceived depth for realistic, colored objects with varying shading was explored. Background: Studies have shown that warm-colored stimuli tend to appear nearer in depth than cool-colored stimuli. The majority of these studies asked human observers to view physically equidistant, colored stimuli and compare them for relative depth. However, in most cases, the stimuli presented were rather simple: straight colored lines, uniform color patches, point light sources, or symmetrical objects with uniform shading. Additionally, the colors were typically highly saturated. Although such stimuli are useful for isolating and studying depth cues in certain contexts, they leave open the question of whether the human visual system operates similarly for realistic objects. Method: Participants were presented with all possible pairs from a set of differently colored objects and were asked to select the object in each pair that appears closest to them. The objects were presented on a standard computer screen, against 4 different uniform backgrounds of varying intensity. Results: Our results show that the relative strength of color as a depth cue increases when the colored stimuli are presented against darker backgrounds and decreases when presented against lighter backgrounds. Conclusion: Color does impact our depth perception even though it is a relatively weak indicator and is not necessarily the overriding depth cue for complex, realistic objects. Application: Our observations can be used to guide the selection of color to enhance the perceived depth of objects presented on traditional display devices and newer immersive virtual environments.

Department of Computer Science & Engineering

2007-18

The Effect of Object Color on Depth Ordering

Authors: Reynold Bailey, Cindy Grimm, Christopher Davoli, Richard Abrams

Corresponding Author: rjb1@cse.wustl.edu

Abstract: The relationship between color and perceived depth for realistic, colored objects with varying shadingwas explored. Background: Studies have shown that warm-colored stimuli tend to appear nearer in depth thancool-colored stimuli. The majority of these studies asked human observers to view physically equidistant,colored stimuli and compare them for relative depth. However, in most cases, the stimuli presented were rathersimple: straight colored lines, uniform color patches, point light sources, or symmetrical objects with uniformshading. Additionally, the colors were typically highly saturated. Although such stimuli are useful for isolating andstudying depth cues in certain contexts, they leave open the question of whether the human visual systemoperates similarly for realistic objects. Method: Participants were presented with all possible pairs from a set ofdifferently colored objects and were asked to select the object in each pair that appears closest to them. Theobjects were presented on a standard computer screen, against 4 different uniform backgrounds of varyingintensity. Results: Our results show that the relative strength of color as a depth cue increases when the coloredstimuli are presented against darker backgrounds and decreases when presented against lighter backgrounds.Conclusion: Color does impact our depth perception even though it is a relatively weak indicator and is notnecessarily the overriding depth cue for complex, realistic objects. Application: Our observations can be used to

Type of Report: Other

Department of Computer Science & Engineering - Washington University in St. LouisCampus Box 1045 - St. Louis, MO - 63130 - ph: (314) 935-6160

Depth by Color

1

Running head: OBJECT COLOR ON DEPTH ORDERING

THE EFFECT OF OBJECT COLOR ON DEPTH ORDERING

REYNOLD J. BAILEY,1 CINDY M. GRIMM,

1 CHRISTOPHER C. DAVOLI,

2 and

RICHARD A. ABRAMS2

1Washington University in St. Louis, Department of Computer Science and Engineering,

St. Louis, Missouri, USA

2Washington University in St. Louis, Department of Psychology,

St. Louis, Missouri, USA

Address correspondence to:

Reynold J. Bailey

Department of Computer Science and Engineering

Campus Box 1045

Washington University in St. Louis

St. Louis, MO, 63130

Email: rjb1@cse.wustl.edu

Depth by Color

2

Abstract

Objective: The relationship between color and perceived depth for realistic, colored

objects with varying shading was explored. Background: Studies have shown that warm-

colored stimuli tend to appear nearer in depth than cool-colored stimuli. The majority of

these studies asked human observers to view physically equidistant, colored stimuli and

compare them for relative depth. However, in most cases, the stimuli presented were

rather simple: straight colored lines, uniform color patches, point light sources, or

symmetrical objects with uniform shading. Additionally, the colors were typically highly

saturated. Although such stimuli are useful for isolating and studying depth cues in

certain contexts, they leave open the question of whether the human visual system

operates similarly for realistic objects. Method: Participants were presented with all

possible pairs from a set of differently colored objects and were asked to select the object

in each pair that appears closest to them. The objects were presented on a standard

computer screen, against 4 different uniform backgrounds of varying intensity. Results:

Our results show that the relative strength of color as a depth cue increases when the

colored stimuli are presented against darker backgrounds and decreases when presented

against lighter backgrounds. Conclusion: Color does impact our depth perception even

though it is a relatively weak indicator and is not necessarily the overriding depth cue for

complex, realistic objects. Application: Our observations can be used to guide the

selection of color to enhance the perceived depth of objects presented on traditional

display devices and newer immersive virtual environments.

Depth by Color

3

THE EFFECT OF OBJECT COLOR ON DEPTH ORDERING

Various sources of information in a scene contribute to our perception of depth.

These sources are referred to as depth cues and are typically grouped into three

categories: pictorial, occulomotor and stereo cues (Gillam, 1995). Pictorial depth cues are

ones that we obtain from two-dimensional images. These include information such as

relative size, distance to horizon, focus, occlusion, shadows, shading, relative brightness,

atmospheric effects, and color (Pfautz, 2000). Perspective cues such as relative size,

occlusion, and distance to the horizon are generally more effective at conveying depth

than shading, luminance, and color (Hone & Davis, 1993; Surdick et al., 1994; Wanger,

Ferwerda, & Greenberg, 1992).

Although color is considered to be one of the weaker depth cues, its impact on

depth perception has been studied extensively by psychologists and other vision

researchers. Artists and designers refer to colors that appear closer to the red end of the

visible spectrum as warm colors, and colors that appear closer to the blue end as cool

colors.1 Early references to the color-depth relationship in literature refer to reds and

oranges as advancing colors and blues as receding colors or retiring colors (Goethe,

1810/1982; Luckiesh, 1918). The actual experimental study of the effect of color on

perceived depth began in latter half of the 19th

century (see Payne, 1964 and Sundet, 1978

for summaries). In most of these studies, different colored stimuli were presented to

human observers and their depth preferences were recorded. It was generally observed

that warmer colors tend to appear closer to the viewer than cooler colors.

In order to account for these observed depth differences, several theories that are

based on the physiology of the human visual system have been proposed. For example, it

Depth by Color

4

has been observed that the color sensitive cones in the retina exhibit a slight bias

(stronger responses) to colors toward the warm end of the visible spectrum (see Figure

1a). Some physiologists have suggested that this bias may be strong enough to result in

perceived depth differences between colors (Livingstone, 2002).

Another (more accepted) theory states that the difference in perceived depth is

due to the fact that shorter wavelengths of visible light are refracted more than longer

wavelengths (Sundet, 1978). As a result, equidistant sources of differing wavelengths

cannot be simultaneously focused onto the retina by the eye's optical system. This is

referred to as chromatic aberration and is illustrated in Figure 1b.

A review of the research literature on color-depth experiments (and in fact most

depth-cue experiments) reveals that in most cases the stimuli presented were very simple.

The stimuli generally consisted of straight colored lines, uniform color patches, point

light sources, or symmetrical objects with uniform shading. Additionally, the colors used

were typically highly saturated.

There are two main reasons for the use of such simple stimuli. First, many early

researchers used the weak observer theory (Landy, Maloney, Johnston, & Young, 1995)

as a basis for their experiments. This theory suggests that the various depth cues are

processed by the visual system independently of each other and our final overall depth

perception is simply a weighted combination of the contribution of the individual depth

cues. Additionally, while most modern researchers quickly dismiss this theory, it is

difficult to design effective experiments for the alternative strong observer theory

(Doorschot, Kappers, & Koenderink, 2001; Guibal & Dresp, 2004; Landy et al., 1995).

The strong observer theory suggests that there are complex interactions between the

Depth by Color

5

various depth cues. The presence of one or more depth cues may either enhance or

suppress the effect of other cues. Secondly, the use of simple stimuli by early researchers

can be attributed to the lack of computing power, display technology, and accurate

measuring devices. While simple stimuli are useful in isolating and studying depth cues

in certain contexts, they leave open the question of whether the human visual system

operates similarly for realistic objects.

The knowledge gained from the study of human depth perception has been used

by computer graphics researchers to develop techniques for manipulating the apparent

depth of objects in computer-generated scenes. These techniques typically rely on three-

dimensional representations of the scene where depth information is readily available.

Examples include the introduction of atmospheric effects like haze or fog (Woo, Neider,

Davis, & Shreiner, 1999), perspective changes such as the shape, size, or position of

objects (or camera) (Foley, van Dam, Feiner, & Hughes, 1996) and level of detail (LOD)

changes (Heckbert & Garland, 1994). In the case of images, where a 3D representation of

the scene is not available, Gooch and Gooch (2004) have shown that the perceived depth

can be enhanced by adding an artistic matting. More recently, Bailey and Grimm (2006)

have shown how subtle apparent depth changes can be introduced by manipulating either

the luminance or color of the object and/or background. The results of the experiment

presented in this paper can be used to guide the selection of color to further enhance the

perceived depth of objects presented on traditional display devices and newer immersive

virtual environments.

This experiment was designed to explore the effect of color on perceived depth

for a realistic colored object with varying shading and contours. Participants were

Depth by Color

6

presented with all possible pairs from a set of differently colored objects and were asked

to select the object in each pair that appears closest to them. This approach is called a

paired comparison test. Ledda, Chalmers, Troscianko, and Seetzen (2005) present a

strong case for the use of this type of testing. Although the number of test pairs (and

hence experiment time) grows quickly as the number of objects in the test set increases,

they point out that the results of a paired comparison test do not suffer from the distortion

that is common in ranking or rating approaches (Siegel & Castellan, 1988). The objects

were presented against 4 different uniform backgrounds of varying intensity levels. This

allows us to analyze the impact of luminance contrast on the depth preferences recorded.

METHOD

Participants

Fifteen students (4 females, 11 males), between the ages of 18 and 40 volunteered

to participate in this study. They all reported normal or corrected-to-normal vision with

no color vision abnormalities.

Stimuli

Stimuli were presented on a 17 inch monitor, operating at 60 Hz with a resolution

of 1280 x 1024. The stimuli used in this experiment consisted of images of 7 colored

teapots and 7 uniform colored patches. To create the stimuli, we began with a photograph

of a teapot2 (see Figure 2a) and performed segmentation using an interactive min-cut

max-flow segmentation algorithm proposed by Boykov and Jolly (2001). Seven

perceptually distinct colors were manually selected (see Figure 2b). The luminance of

these colors was adjusted to match the average luminance of the original teapot. The

resulting equiluminant colors are shown in Figure 2c. Each of these target colors was

Depth by Color

7

then applied to the teapot. This was done by scaling the target colors by the pixel

luminance values of the original teapot. This approach preserves the shading and specular

highlights of the original teapot (see Figure 3a). The 7 colored uniform patches were

simply images of each of the equiluminant target colors (see Figure 3b). Each teapot

image and each color patch image subtended a visual angle of 5o (i.e. 5

o high and 5

o

wide).

Procedure and Design

Participants were seated in front of a computer screen in a dimly lit room. Two

differently colored objects of the same type (teapot or color patch) were presented on

screen for each trial. The objects were displayed on either side of an imaginary central

line on the screen. The background for both objects in a single trial could be one of the

four different uniform levels of gray, evenly spaced from black to white. Participants

were instructed to decide which object appeared nearer and responded by moving the

mouse over their choice and clicking on it. If participants were unsure, they were

instructed to choose based on their initial impulse. Participant response time for each trial

was also recorded. Between each trial, a blank screen was displayed for 2 seconds. This

was done because it has been shown that our perception of depth diminishes with viewing

time (Troscianko, Montagnon, Le Clerc, Malbert, & Chanteau, 1991). A 2 second

interval prevents total adaptation, thereby ensuring that depth perception remains stable

(Troscianko et al., 1991). Each participant performed 21 comparisons for each type of

object against each of the 4 possible backgrounds, for a total of 168 experimental trials.

An object of a particular color was never compared against itself, but all other within-

object color comparisons occurred. All trials that fell under one type of object and one

Depth by Color

8

type of background comprised a block, for a total of eight blocks (i.e. teapots presented

against each of the 4 backgrounds, and color patches presented against each of the 4

backgrounds). The presentation order of the blocks was randomly determined and trials

within each block were randomly ordered. It was randomly determined for each trial

which object would be on the left and which would be on the right. Participants were

given breaks after every 7 minutes of testing.

Scoring

Participant responses were stored in 7 x 7 preference matrices (see Figure 4a).

This approach (patterned after Ledda et al., 2005) allows for easy analysis of the results.

Eight preference matrices were generated for each participant: four for the teapot data set

presented against each of the 4 possible backgrounds and likewise four for the color patch

data set. A preference matrix is best interpreted in a row-wise fashion. For example in

Figure 4a, the 1's in the first row specify that the purple teapot appears closer than the

orange and red teapots.

RESULTS

Corresponding preference matrices of all participants were summed together to

form eight cumulative preference matrices. Figure 4b shows the cumulative matrix for

the 15 participants for the teapot data set presented against the black background.

Summing across each row produces a score for each colored teapot (shown in last

column). This score tells how many times a given teapot was chosen as “nearer” relative

to all other teapots. Figure 5 shows plots of the scores for each teapot and color patch

against each of the 4 backgrounds. The two darker uniform backgrounds were darker than

the average luminance of the stimuli while the two lighter uniform backgrounds were

Depth by Color

9

lighter than the average luminance of the stimuli. Notice that when stimuli were

presented against the darker backgrounds that a clear cyclic trend emerged in the scores

from cool to warm. This cyclic behavior correlates nicely with the familiar circular

representation of the color spectrum.3

Notice however that as the background luminance

increased that this cyclic relationship was lost. This suggests that the relative strength of

color as a depth cue is enhanced against darker backgrounds.

Figure 6 shows plots of all the recorded response times as a function of variance

from the expected mean probability (0.5). The red line in each plot illustrates the best

linear fit of the data. Guibal and Dresp (2004) observed a correlation between the

probability of a “near” response and the associated response time (the higher the

probability, the shorter the response time) for simple highly saturated stimuli. Such

observations are often used to help validate experimental setup. We observed similar

trends in our experiment as illustrated in Figure 6.

Figure 7 and Figure 8 show the average probability of responding “near” for each

of the teapots and color patches over the 4 possible backgrounds. The trends in the

average probability of responding “near” for each color patch (Figure 7) revealed less

variability than the corresponding trends for the teapots (Figure 8). This is evidenced by

the relatively smooth functions for the color patches. Note that for the color patches, red,

which has been established in the literature as appearing nearer, does indeed show this

pattern: the red patch was chosen reliably more than the purple, cyan, green, yellow and

orange patches. In the teapots however, the red teapot was chosen reliably more than the

cyan, green, and yellow teapots only. Additionally, the warm orange color which

Depth by Color

10

received a high probability of being reported as “near” for the color patches was less

effective when applied to the teapot.

DISCUSSION

The present study was conducted to explore the effect of color on the perceived

depth of realistic objects. Previous studies using simple stimuli have been useful in

isolating and studying depth cues in certain contexts, but have left open the question of

whether the human visual system operates similarly for realistic objects. The results

presented here suggest that the more complex an image is, the more an observer may try

to gather multiple depth cues from that image, supporting the strong observer theory

(Doorschot, Kappers, & Koenderink, 2001; Guibal & Dresp, 2004; Landy et al., 1995).

When complex images are viewed, such as the teapots, color is not necessarily the

overriding depth cue. Additionally, the results show that there are circumstances under

which the relative strength of color as a depth cue can be enhanced. For both the simple

patches and the more complex teapots, color functioned as a more powerful depth cue

against darker backgrounds than against lighter backgrounds. These results provide

valuable insight into the impact of luminance contrast on our depth perception.

The observations made in this experiment have several useful applications to the

field of Computer Graphics. First, many modern software systems such as 3D-modeling

programs, paint programs, or even word processors and spreadsheets have a plethora of

user options available. This often leads to complex user interfaces with numerous

buttons, sliders, and drop-down menus in multiple panels. The clever use of non-

perspective depth cues such as color may further enhance the readability and hence user-

friendliness of such complex user-interfaces. By incorporating our observations during

Depth by Color

11

the user-interface design stage of the software development process, suitable colors can

be selected to ensure that these goals are met. Second, several recent studies have shown

that humans generally underestimate distance and depth4

in virtual environments (Loomis

& Knapp, 2003; Thompson, Willemsen, Gooch, Creem-Regehr, Loomis, & Beall, 2004;

Willemsen & Gooch, 2002). Our observations can be used to guide the selection of color

to help improve the perceived depth of objects displayed on such systems. Finally,

consider an electronic visual alert or warning system that demands quick attention. Our

results (in addition to that of Guibal and Dresp (2004)) show that certain foreground /

background color combinations result in stronger depth cues and that participants

responded quicker to these cues. By making informed color selection decisions, an

additional depth cue can be introduced which might bring the alert into awareness

quicker than an otherwise equal alert that did not capitalize on this cue.

Depth by Color

12

References

Boykov, Y. & Jolly, M.-P. (2001). Interactive graph cuts for optimal boundary and region

segmentation of objects in N-D images. Eighth International Conference on

Computer Vision, 1, 105.

Bailey, R. J., & Grimm, C. M. (2006). Perceptually meaningful image editing: Depth

(Tech. Rep. No. 2006-11). St. Louis: Washington University, Department of

Computer Science and Engineering.

Doorschot, P. C. A., Kappers, A. M. L., & Koenderink, J. J. (2001). The combined

influence of binocular disparity and shading on pictorial shape. Perception and

Psychophysics, 63, 1038-1047.

Foley, J. D., van Dam, A., Feiner, S. K., & Hughes, J. F. (1996). Computer graphics –

principles and practice (Second edition in C). Boston: Addison-Wesley.

Gillam, B. (1995). The perception of spatial layout from static optical information. In W.

Epstein & S. Rogers (Eds.), Perception of space and motion (p. 23-67). San

Diego: Academic Press.

Goethe. (1982). Theory of colours (C. L. Eastlake, Trans.). Cambridge, Massachusetts:

M.I.T. Press. (Original work published 1810)

Gooch, A. A., & Gooch, B. (2004). Enhancing perceived depth in images via artistic

matting. In Proceedings of the 1st symposium on applied perception in graphics

and visualization (p. 168). New York: ACM Press.

Guibal, C. R. C., & Dresp, B. (2004). Interaction of color and geometric cues in depth

perception: When does red mean near? Psychological Research, 69, 30-40.

Heckbert, P., & Garland, M. (1994). Multiresolution modeling for fast rendering. In

Depth by Color

13

Proceedings of Graphics Interface (pp. 43-50). Banff, Alberta, Canada: Canadian

Information Processing Society.

Hone, G., & Davis, R. (1993). Brightness and depth on the flat screen: Cue conflict in

simulator displays. SPIE Proceedings – Human Vision, Visual Processing, and

Digital Display IV, 1913, 518-528.

Landy, M. S., Maloney, L. T., Johnston, E. B., & Young, M. (1995). Measurement and

modeling of depth cue combination: In defense of weak fusion. Vision Research,

35, 389-412.

Ledda, P., Chalmers, A., Troscianko, T., & Seetzen, H. (2005, August). Evaluation of

tone mapping operators using a high dynamic range display. Proceedings of ACM

SIGGRAPH 2005, 24, 640-648.

Livingstone, M. (2002). Vision and art: The biology of seeing. New York: Harry N.

Abrams, Inc.

Loomis, J. M. & Knapp, J. M. (2003). Visual perception of egocentric distance in real

and virtual environments. In L. J. Hettinger & M. W. Haas (Eds.), Virtual and

Adaptive Environments (p. 21-46). Mahwah, NJ: Lawrence Erlbaum Associates.

Luckiesh, M. (1918). On retiring and advancing colors. American Journal of Psychology,

29, 182-186.

Newton, I. (1966). Opticks. Brussels, Belgium: Culture et Civilisation. (Reprinted from

Opticks, printed for S. Smith and B. Walford, 1704, London)

Payne, M. C., Jr. (1964). Color as an independent variable in perceptual research.

Psychological Bulletin, 61, 199-208.

Pfautz, J. (2000). Depth perception in computer graphics. Doctoral dissertation,

Depth by Color

14

University of Cambridge, Cambridge, UK.

Siegel, S., & Castellan, N. (1988). Nonparametric statistics for the behavioral sciences

(2nd

ed.). New York: McGraw–Hill, Inc.

Stockman, A., & Sharp, L. T. (2000). Spectral sensitivities of the middle- and long-

wavelength sensitive cones derived from measurements in observers of known

genotype. Vision Research, 40, 1711-1737.

Stockman, A., Sharp, L. T., & Fach, C. C. (1999). The spectral sensitivity of the human

short-wavelength cones. Vision Research, 39, 2901-2927.

Sundet, J. M. (1978). Effects of colour on perceived depth: Review of experiments and

evaluation of theories. Scandinavian Journal of Psychology, 19, 133-143.

Surdick, R., Davis, E., King, R., Corso, G., Shapiro, A., Hodges, L., et al. (1994).

Relevant cues for the visual perception of depth: Is it where you see where it is?

HFES 38th Annual Meeting, 1305-1309.

Troscianko, T., Montagnon, R., Le Clerc, J., Malbert, E., & Chanteau, P. (1991). The role

of colour as a monocular depth cue. Vision Research, 31, 1923-1930.

Thompson, W. B., Willemsen, P., Gooch, A. A., Creem-Regehr, S., Loomis, J. M., &

Beall, A. C. (2004). Does the quality of computer graphics matter when judging

distances in visually immersive environments? Journal of Presence:

Teleoperators and Virtual Environments, 13, 560-571.

Wanger, L., Ferwerda, J., & Greenberg, D. (1992). Perceiving spatial relationships in

computer-generated images. Computer Graphics and Applications, 12, 44-58.

Willemsen, P. & Gooch, A. A. (2002). Perceived egocentric distances in real, image-

Depth by Color

15

based, and traditional virtual environments. In VR ’02: Proceedings of the IEEE

Virtual Reality Conference (p. 275). Washington, DC: IEEE Computer Society.

Woo, M., Neider, J., Davis, T., & Shreiner, D. (1999). OpenGL Programming Guide (3rd

ed.). Reading, MA: Addison-Wesley.

Depth by Color

16

Footnote

1 Physicists use the opposite terminology: on a black-body radiator, bluer colors

are said to be hot while redder colors are said to be cool.

2 This particular teapot was chosen because it appeared natural regardless of its

color.

3 The color spectrum is often presented as a continuous color wheel with red

blending into purple. The original color wheel is credited to Sir Isaac Newton

(1704/1966).

4 The terms distance and depth are often incorrectly used interchangeably. In the

context of virtual environments, distance can either be defined to be egocentric (distance

from self) or exocentric (relative distance between objects). Depth refers to egocentric

distance.

Depth by Color

17

Figure Captions

Figure 1. (a) Response of the long, medium, and short wavelength (L, M, and S) sensitive

cones to different wavelengths of light. Data based on experiments conducted on human

observers (Stockman & Sharp, 2000; Stockman, Sharp, & Fach, 1999). (b) Chromatic

aberration. Assumes eye is red accommodated. Adapted from Sundet (1978).

Figure 2. (a) Original teapot photograph. Courtesy of Jack Tumblin and Ankit Mohan.

(b) Manually selected perceptually distinct colors. (c) Adjusted equiluminant colors.

Figure 3. Stimuli used in experiment: (a) teapots and (b) uniform color patches.

Figure 4. (a) Preference matrix. A 1 in row i, column j signifies that stimulus i appeared

closer than stimulus j. (b) Cumulative matrix for teapot data set against black

background.

Figure 5. (a) Cumulative scores for each teapot against each of the 4 backgrounds. (b)

Cumulative scores for each color patch against each of the 4 backgrounds.

Figure 6. (a) Participant response times for teapot data set against each of the 4

backgrounds as a function of variance from the expected mean probability (0.5). (b)

Participant response times for color patch data set against each of the 4 backgrounds as a

function of variance from the expected mean probability (0.5). The red line in each plot

illustrates the best linear fit of the data.

Figure 7. Average probability of responding “near” for each of the teapots over the 4

possible backgrounds.

Figure 8. Average probability of responding “near” for each of the color patches over the

4 possible backgrounds.

Depth by Color

18

Figure 1

Depth by Color

19

Figure 2

Depth by Color

20

Figure 3

Depth by Color

21

Figure 4

Depth by Color

22

Figure 5

Depth by Color

23

Figure 6

Depth by Color

24

Figure 7

Depth by Color

25

Figure 8

Depth by Color

26

Biographies

Reynold J. Bailey

Department of Computer Science and Engineering, Washington University in St. Louis

M.S. Computer Science, 2004, Washington University in St. Louis, USA

Cindy M. Grimm

Department of Computer Science and Engineering, Washington University in St. Louis

Ph.D. Computer Science, 1996, Brown University, Rhode Island, USA

Christopher C. Davoli

Department of Psychology, Washington University in St. Louis

B.S. Psychology, 2004, Davidson College, North Carolina, USA

Richard A. Abrams

Department of Psychology, Washington University in St. Louis

Ph.D. Psychology, 1986, University of Michigan, USA