Partial Exclusion of the Local Effect from the ... - Springer Link

OPTICAL REVIEW Vol. 13, No. 5 (2006) 380–387

Partial Exclusion of the Local Effect from the Assessment of Recognized Illuminant Using Depth Separation Kitirochna R ATTANAKASAMSUK and Hiroyuki SHINODA Human Vision Laboratory, Graduate School of Science and Engineering, Ritsumeikan University, Kusatsu, Shiga 525-8577, Japan (Received November 7, 2005; Accepted July 27, 2006) Perceptual achromatic setting has been claimed to be a method of assessing the recognition of illumination in an environment. The result, however, is not spatially valid because of some effect of an immediate area of the test (local effect). We considered a method for excluding this local effect from the assessment. We hypothesized base on the coplanar theory and the recognized visual space of illumination theory that a depth separation between the test and its immediate area would be an effective way. In the first experiment, the perceptual achromatic setting was conducted in a natural scene, with several test luminance levels. The perceptual achromatic locus from two conditions was compared. One condition is the test pasted on a chromatic surround. The other is the test fronto-parallelly separated from the surround. In the second experiment, the perceptual achromatic setting was conducted at the several depth separations. We found that the depth separation decreases the local effect but it was not completely excluded. # 2006 The Optical Society of Japan Key words: recognition of the illumination, local effect, chromatic induction, depth separation, depth perception

1.

local effect significantly affects the perceptual achromatic locus in some case.7–9) Therefore, the perceptual achromatic locus does not validly represent the chromaticity of the recognized illuminant unless the local effect is completely excluded. In this study, we questioned whether the local effect can be excluded or not. Either co-planar theory10) or RVSI theory4) has claimed that the plane or space, where the test is belonging to, show an influence on judgment of the appearance of the test. If belongingness of the test to its immediate area is destructed, we expect the local effect would be vanished. We thought fronto-parallel separation of the test from its immediate area would be an effective way to destruct the belongingness without changing retinal image of the stimulus. We call this method as ‘‘depth separation’’. We then investigated whether the depth separation is an effective method for excluding the local effect or not. We built an experimental room illuminated with white illumination. A variable color test and a replaceable local surround were placed in this room. Subject’s task was to make the perceptual achromatic setting. The results were compared among three experimental conditions. In the first condition, the test was presented as if it was pasted on a neutral local surround. This condition was a control condition. The perceptual achromatic locus of this condition was a reference locus for this experiment. In the second condition, the experimental setup was the same as the setup in the first condition except that the neutral surround was replaced by a chromatic surround. This condition would exhibit the existence of the local effect. We expected the perceptual achromatic locus of the second condition would shift toward the chromaticity of the chromatic surround. In the third condition, the experimental setup was the same as the setup in the second condition except that the depth separation was employed. The test was fronto-parallelly separated from the chromatic surround without changing the

Introduction

Many pieces of research have shown that physical illumination is not a reference of visual system for perceiving appearance of a stimulus. On the contrary, some representation that visual system constructs for understanding the state of illumination such as perceived illuminant,1) inferred illuminant,2) equivalent illuminant3) and recognized visual space of illumination (RVSI)4) is used for determining appearance of the stimulus. We call the representation as recognized illuminant. The appearance of stimulus in visual scene is supposed to be determined in relation to the color of the recognized illuminant. In order to assess the color of the recognized illuminant a method called perceptual achromatic setting is generally used. This method is based on the following assumption that if a physical achromatic patch is placed under any illuminations, color of that patch then results from the color of illumination. By the same way, we assumed that perceived color of the physical achromatic patch results from the color of recognized illuminant. Once we cancel the perceived color on the patch until it appears achromatic, that amount will therefore represent the effect of recognized illumination. And perceptual achromatic locus of that patch is claimed to be chromaticity of the recognized illuminant. However, it is also well known that appearance of stimulus also depend on immediate area surrounding the stimulus.5) We call the effect of immediate area as a local effect. A well known phenomenon exhibiting the local effect is chromatic induction or sometime called simultaneous color contrast. A gray patch appears, for instance, reddish when it is surrounded by a green color. Some research has claimed that magnitude of the local effect is small and the local effect slightly shows an influence on the perceptual achromatic locus.3,6) Nonetheless, the magnitude of the local effect is substantial and the 380

OPTICAL REVIEW Vol. 13, No. 5 (2006)

K. R ATTANAKASAMSUK and H. SHINODA

subject’s retinal image. If the depth separation is an effective way to exclude the local effect, the perceptual achromatic locus of the third condition should differ from that of the second condition and coincide with the reference locus. If the local effect could not exclude by the depth separation, there should be no difference of the perceptual achromatic locus between the second and third condition. 2.

Experiment 1

2.1 Method The apparatus was a 75 140 120 cm3 (W L H) box decorated with wall paper as shown in Fig. 1. Some objects such as a Macbeth Color Checker Chart and a doll were put into this box. The box was illuminated by a set of 6500 K fluorescence lamp (FL). The illumination intensity was at 440 lx measured on the floor directly under FL with Minolta CL-200 illuminometer. Chromaticity of the illumination was ð0:188; 0:480Þ in CIE u0 v0 . A 10 10 replaceable Munsell paper was pasted on the front panel, treated as a local surround (S). The local surround was three kinds (Sn, Sg, and Sb) as their properties were shown in Table 1. A test (T) was a Munsell N4.5 paper hung on a stand attached to the liquid crystal display (LCD) projector (LCD) holder. The test is tilted at 45 angle from vertical plane facing downward to minimize the light from FL light. The chromaticity and luminance of the test was controlled by projecting light from LCD. This set of test stimulus was movable so that it could be placed at any position. In this experiment, it was mainly composed of two conditions for the test position; where the test was presented on same plane of the local surround (‘‘Wall’’ condition) and presented on the different plane in front of the local surround (‘‘Midair’’ condition). In the Wall condition, the set of test stimulus was placed behind the box. A 1:5 1:5 hole was made at the center of the surround. A subject saw the test

120 cm

20 cm

FL

S 70 cm

T

120 cm

FL

S T

LCD

(a)

(b)

Fig. 1. (a) Schematic of the apparatus. (b) The subject’s view. FL, the fluorescent lamp; S, the replaceable local surround; T, the test (N4.5 Munsell paper); LCD, The LCD projector: solid line shows the position of the LCD in the Wall condition, dashed line shows the position of the LCD in the Midair condition.

Table 1. Properties of local surrounds. Local surround

CIE u0 v0 chromaticity

Luminance (cd/m2 )

Sg (5G6/9) Sb (10B6/9) Sn (N6)

ð0:141; 0:512Þ ð0:156; 0:422Þ ð0:198; 0:476Þ

37 36 37

381

through the hole with binocular viewing. The N4.5 Munsell paper was large enough so that the subject could not see the edge of the N4.5 patch. Therefore, the subject’s retinal image between the left and right eye were identical. No binocular disparity resulted in an appearance of the test pasted on the same plane of the local surround. In the Midair condition, the depth separation between the test and the local surround was employed. The set of test stimulus was placed inside the box (shown by the dashed line in Fig. 1). The test was presented at 50 cm in front of the center of the local surround. Due to the binocular disparity, the subject perceived the different depth between the test and the local surround. The test size was kept constant at 1:5 1:5 so that the subject’s retinal image is the same as that in the Wall condition. The set of stimulus was carefully calibrated at each condition with a Minolta CS-100 Chroma Meter. Matlab and Psychophysics Toolbox for Windows Extensions11) were used to generate computer programs for calibration and running the experiment. Two male subjects with normal color vision participated in this experiment. Each session began with two min adaptation in the experimental room. They were instructed to look around in the experimental room during the adaptation. After two min adaptation, the stimulus was presented. The starting chromaticity and luminance of the test were randomly selected. Subject’s task is to observe the test with binocular viewing and to adjust the appearance of the test with a trackball until the test appears achromatic. During each setting, the luminance of the test was kept constant. The subject was instructed to look around in the experimental room and not to fix their eyes at the test. There was no time limit for each setting. When the subject achieved satisfactory adjustment, chromaticity of the test was measured at the subject’s position using the Minolta CS100 Chroma Meter. After recording the measured data, the next stimulus was presented. Within each session, the local surround and the test’s position were fixed but eight test luminance levels at 23, 30, 37, 55, 70, 95, 170, and 340 cd/m2 were randomly presented. Each test luminance level was presented 3 times making a total 24 settings per session. Each subject has done three sessions per experimental condition. The experimental conditions were composed of the combination of three local surrounds (Sn, Sg, and Sb) and two test’s positions (Wall and Midair). Each subject made total 432 setting in this experiment. 2.2 Result and discussion The result obtained from subject KR setting on Sg with the test luminance level at 37 cd/m2 is plotted in a part of CIE u0 v0 chromaticity diagram as shown in Fig. 2. Cross, gray circle, and black circle represent chromaticity of FL, Sn, and Sg, respectively. Gray triangle (Sn-Wall) and open triangle (Sn-Midair) represent the chromaticity of the perceptual achromatic loci setting on Sn in the Wall and the Midair condition, respectively. We assume that the neutral surround will not cause a chromatic induction. The appearance of the test would be entirely determined by the recognized illuminant. Therefore, Sn-Wall and Sn-Midair

382



tion (FL). It has been shown that the understanding of illumination is unnecessary to be related with the physical illuminant.12) Black square (Sg-Wall) and open square (Sg-Midair) represent the chromaticity of the perceptual achromatic loci setting on Sg in the Wall and the Midair condition, respectively. The result shows that Sg-Wall is not coincided with Sn-Wall and shifts toward the chromaticity of Sg. This result agrees with the previous work7–9) that the local effect shows an influence on the assessment of recognized illuminant. When test and its chromatic local surround are fronto-parallelly separated, Sg-Midair does not coincide with Sg-Wall but shifts toward Sn-Wall. However, the shift of Sg-Wall is not large enough to coincide with Sn-Wall. From this result, there are some points should be noted here. First, the difference between Sg-Wall and Sg-Midair confirms the important role of depth on color appearance. Second, the shift of Sg-Wall toward Sn-Wall indicates that the local effect is diminished by the depth separation. Both points agree well with previous work which exhibited that color appearance is affected by depth perception. Third, the insufficient shift of Sg-Wall indicates that the local effect is still exist and not completely excluded by the depth separation. Figure 3 shows all result of both subjects. We found the perceptual achromatic loci are located along the line between Sn and the chromatic local surround (Sg and Sb). Therefore, we present only one dimension of the loci in order to make the relationship between the perceptual

0.54

Sg-Wall Sg-Midair Sn-Wall Sn-Midair Sg Sn FL

CIE v’

0.51

0.48

0.45 0.12

0.15 CIE u’

0.18

0.21

Fig. 2. Perceptual achromatic loci of subject KR (test luminance level at 37 cd/m2 ) plotted in CIE u0 v0 chromaticity diagram. Cross, black circle, and gray circle represent the chromaticity of FL, Sg, and Sn, respectively. Black and opened square represent the perceptual achromatic loci of setting on Sg in the Wall and the Midair position, respectively. Gray triangle and open triangle represent the perceptual achromatic loci setting on Sn in the Wall and the Midair position, respectively. Error bars show SD.

Perceptual achromatic loci (CIE u’)

could represent the chromaticity of the recognized illuminant. These two loci almost coincide with each other and we use Sn-Wall as a reference in the later discussion. Moreover, there is another reason that this Sn-Wall is more suitable to be the reference than the chromaticity of physical illumina-

0.13

KR

0.14

0.14

0.15

0.15

0.16

0.16

0.17

0.17

0.18

0.18

0.19

0.19

0.20

0.20

PC

Sn-Wall Sg-Wall Sg-Midair Sb-Wall Sb-Midair Sg, Sb Sn FL

0.21

0.21 10

100

0.15


0.13

10

1000

KR

0.16

0.16

0.17

0.17

0.18

0.18

0.19

0.19

0.20

0.20

0.21

100

0.15

1000

PC

0.21 10

100 Test luminance (cd/m 2)

1000

10

100 Test luminance (cd/m 2 )

1000

Fig. 3. CIE u0 chromaticity of the perceptual achromatic loci is plotted on the vertical axis against the test’s luminance level plotted in log scale on the horizontal axis. (See detail in text.)



Reduction ratio ¼ A=B

ð1Þ

Where A is size of vector between Sg-Wall (or Sb-Wall) and Sg Midair (or Sb-Midair) and B is size of vector between SgWall (or Sb-Wall) and Sn-Wall.

1.0

KR

Reduction ratio

0.8

0.6

0.4

0.2

0.0 1.0

10

100

1000

PC

0.8

Reduction ratio

achromatic loci and the test luminance level clearer. CIE u0 of the perceptual achromatic loci was plotted in reverse order on the vertical axis against the test’s luminance plotted in logarithm scale on the horizontal axis. Each point represents the average of nine settings. Gray square (SnWall), black square (Sg-Wall), and black triangle (Sb-Wall) represent CIE u0 chromaticity of the perceptual achromatic loci setting on Sn, Sg, and Sb in the Wall condition respectively. Open square (Sg-Midair) and open triangle (Sb-Midair) represent CIE u0 chromaticity of the perceptual achromatic loci setting on Sg and Sb in the Midair condition respectively. Gray and dashed line represent CIE u0 chromaticity of Sn and FL, respectively. Black line in the top and bottom panel represents CIE u0 chromaticity of Sg and Sb, respectively. In Fig. 3, Sn-Wall is quite constant and independent from the test luminance level change. This result is expectable because we assumed that Sn should not cause chromatic induction. The magnitude of the local effect in this condition should be zero. Color appearance of the test should be determined solely by the recognized illuminant. Therefore, the perceptual achromatic loci would not be affected by variation of the test luminance level. When the neutral local surround was replaced with a chromatic one, the perceptual achromatic loci shift toward the chromaticity of the surround. Sg-Wall and Sb-Wall at every test luminance level shows the similar trends of the result. All of them do not coincide with Sn-Wall and shift toward the chromaticity of the local surround. However, the shift of Sg-Wall and Sb-Wall, which can be implied as the magnitude of the local effect, is not a constant but depends on the test’s luminance level. There is a certain test luminance level which the magnitude of the local effect is maximal. If the test luminance level is higher or lower than this level, the magnitude tends to decrease. The luminance dependency of the local effect is similar to that in previous study on the amount of chromatic induction. Bergstrom et al.13) have shown that the highest magnitude of chromatic induction of simple center/surround configuration occurs at the luminance ratio nearer to 1. However, it is not necessary that the luminance ratio solely determined the magnitude but also some factor such as the understanding of illumination might involve too.12,14,15) When the depth separation was employed in the perceptual achromatic setting, Sg-Midair and Sb-Midair differ respectively from their corresponding Sg-Wall and Sb-Wall. Both Sg-Midair and Sb-Midair tend to shift toward Sn-Wall in every test luminance level. This shift indicates the magnitude of reduction of local effect in the Midair condition. Note that the reduction of local effect by the depth separation also depends on the test luminance level. We have calculated the reduction ratio at each test luminance level by the following equation.

383

0.6

0.4

0.2

0.0 10

100

1000

Test luminance (cd/m2 )

Fig. 4. Reduction ratios of the local effect are plotted on the vertical axis against the test’s luminance level plotted in log scale on the horizontal axis. Black square and gray triangle represent the reduction ratio on Sg and Sb, respectively. Each line represents the corresponding regression.

Figure 4 shows the relationship between the reduction ratio and the test luminance level. The reduction ratio at high test luminance level is smaller than at low test luminance level. We are not sure about the cause of this characteristic. A possible explanation might be related to mode of appearance of the test. If the luminance of the local surround is kept constant, increment of the test luminance level change the mode of appearance of the test from object color mode (reflected surface) to be light source (aperture) color mode. In our daily life, a reflected surface can be easily judged the position where it is placed in the scene. On the contrary, sometime we feel difficult in localization of aperture color in depth direction. For example, a floating phenomenon presented by Aoki et al.16,17) is an evidence of perceptual mislocation of aperture color on the object background. In this experiment, both subjects have reported that aperture color is perceived when the test luminance level is beyond 37 cd/m2 . It might be possible that when appearance of the test changed to be aperture color, localization of the test on its local surround in the Wall condition was changed. Perceptual position of the test in the Wall condition possibly appears as if it was floated on the local surround. Even though the real position is different but perceived depth, which occurs in the Wall condition, may cause the similar perceptual position of the test. The depth separation, therefore, shows little influence on the perceptual achromatic setting at high test luminance level.


3.

Experiment 2

In this experiment, the relationship between local effect and the depth separation was investigated. The perceptual achromatic setting was done at several depth separation distances. 3.1 Method The apparatus and experimental procedure was similar to those in previous experiment except the following differences. The position of the set of stimulus was varied, resulting in the distance between the test and its local surround (depth separation distance) of 0, 8, 25, 50, and

70 cm. The experimental setup of 0 and 50 cm distance was respectively analogous to the setup in the Wall and the Midair condition in the previous experiment. The experimental setup of 8, 25, and 70 cm was similar to the setup in the Midair condition except the depth separation distance was changed. In all condition, the test size was kept constant at 1:5 1:5 . Subject made the perceptual achromatic setting with constant test’s luminance level at 37 cd/m2 . The experimental conditions were composed of the combination of three local surrounds and five depth separation distances. Within each session, subject made 3 settings in each experimental condition making a total 45 settings per session. Each subject participated in three sessions for completing the experiment. One male and one female subject with normal color vision were participated in this experiment. 3.2 Result and discussion Figure 5 shows relationship between the perceptual achromatic setting and the depth separation. Gray square (Sn-A) shows CIE u0 chromaticity of the perceptual achromatic loci settings on Sn. Black square (Sg-A) and

0.15


To compare with previous work, our result agrees with similar study related to depth perception and color appearance. Shevell and Millner18) have shown small decremental chromatic induction when surround and test were perceived in different depth plane. Schirillo and Shevell19) have shown that brightness and lightness of a test patch affected by the perceived depth of the test. Shevell and Wei20) have shown that strongest attenuation of the local effect occur when the remote chromatic contrast are on the same depth plane as the remote region. Yamauchi and Uchikawa21) have shown that mode of appearance of a surface is mainly determined by the information which shared the same depth with that surface. However, these works have been conducted on the CRT simulated environment and their amount of simulated depth is quite small. Our experiment was conducted in the real three-dimensional space and used the large amount of real depth. We want to investigate whether the large amount of real depth is possible to completely exclude the local effect from the assessment of the recognized illuminant in the real environment or not. To conclude from the first experiment, the present result partially supports our hypothesis because Sg-Midair and Sb-Midair did not coincide with Sn-Wall. This means the local effect is partially diminished by the depth separation but the local effect still remains. If we considered the previous work, none of them have shown whether the entire local effect is depth dependent or not because only the certain depth is used in those experiments. We then conducted the next experiment to investigate the relationship between the depth separation distance and the amount of the local effect. If the entire local effect is depth dependent, variation of the depth separation distance should alter the magnitude of the local effect. Increasing in the depth separation distance would decrease the magnitude of the local effect. Then, the residual amount of local effect in the first experiment would be due to the insufficient the depth separation distance. Therefore, it is possible to estimate the required depth separation distance where the local effect is vanished. On the other hand, if the only some portion of local effect is depth dependent, the depth separation would show an influence on the reduction of the magnitude of the local effect only at some certain range of depth. Increasing in the depth separation distance beyond this range would not affect the reduction of local effect. Therefore, we cannot estimate the required depth separation distance for excluding the entire local effect.


KR

0.16 0.17 0.18 0.19 0.20 0.21 0

20

40

60

0.15


384

80

YT

0.16 0.17 0.18 0.19 0.20 0.21 0

20

40

60

80

Depth separation (cm)

Sg-A

Sb-A

Sn-A

Fig. 5. CIE u0 chromaticity of the perceptual achromatic loci is plotted in reverse order on the vertical axis against the depth separation distance on the horizontal axis. Gray, black, and open square represent the result from Sn, Sg, and Sb local surround condition, respectively. Error bars show SD. For subject KR, the triangle symbols represent the result from the Wall and the Midair condition in the first experiment.



120 cm

20 cm FL

FL

120 cm

S

S T

T LCD

LCD

(a)

(b)

Fig. 6. (a) Schematic of the apparatus in supplementary experiment. (b) The subject’s view. Labels are the same as used in Fig. 1.


0.15

KR 0.16 0.17 0.18 0.19 0.20 0.21 0


open square (Sb-A) show CIE u0 chromaticity of the perceptual achromatic loci settings on Sg and Sb, respectively. For subject KR, who participated in the first experiment, triangle represents the perceptual achromatic loci setting from the Midair and the Wall condition in the previous experiment. It shows the consistent criterion of subject KR for the perceptual achromatic setting task between two experiments. The results from both subjects show the same trend that Sn-A slightly change even though the depth separation distance is varied. The coincidence of all Sn-A agrees with the result of the first experiment (shown by Sn-Wall and SnMidair in Fig. 1) that the perceptual achromatic loci setting on the neutral surround are not affected by depth between the test and the local surround. For the setting on the chromatic surround condition (Sg-A and Sb-A), the results are twofold. First, all perceptual achromatic loci from the depth separation condition (8, 25, 50, and 70 cm) do not coincide with the perceptual achromatic loci from no depth separation condition (0 cm). Once the depth separation was employed, Sg-A and Sb-A shift toward Sn-A. Second, the perceptual achromatic loci are almost the same among the depth separation conditions (8, 25, 50, and 70 cm). Variation of the depth separation distance slightly affects the shift of the perceptual achromatic loci. From the result mentioned earlier, one might argue that the difference result between two conditions may be caused by the different way of stimulus presentation. In the no depth separation condition, subject see only the test and surround while in the depth separation condition subject see the rod supporting the test. All the perceptual achromatic loci from the depth separation condition show no significant difference possibly because of the same way of stimulus presentation. We then conducted a supplementary experiment to examine the possibility of this artifact. The experimental set up was modified as shown in Fig. 6. The surround was inclined 45 from the front wall to be parallel to the test plane. In the no depth separation condition, the position of the projector set was moved all the way back until the test attached to the surround. Now subjects saw the rod attaching to the test and the projector in both conditions, the depth separation and no depth separation condition. All the properties were the same as those in the experiment 2. Figure 7 shows comparison of the result between the supplementary experiment (triangle) and the experiment 2

385

20

40

60

0.15

80

UT

0.16 0.17 0.18 0.19 0.20 0.21 0

20

40

60

80

Depth separation (cm)

Sg-A

Sb-A

Sn-A

Sg-45

Sb-45

Sn-45

Fig. 7. Comparison of the result from the experiment 2 (square) and supplementary experiment (triangle).

(square). It shows that the result from both experiment are almost the same. This indicates that the difference between the depth and no depth separation in the experiment 2 was not caused by the different way of presenting stimulus. Moreover, two more depth separation distance (1 and 4 cm) was employed to investigate the clearer relationship between the magnitude of local effect and the depth separation distance. The result shows that there is a range where the perceptual achromatic loci are affected by depth. From the result in Fig. 7, the relationship between the magnitude of the local effect and the depth separation distance can be illustrated in Fig. 8. Increasing in the depth separation distance (D) make the magnitude of the local effect (L) decreased. Once the depth separation distance reaches a certain distance, the magnitude of the local effect is not further decreased. We named this certain distance as a critical depth (Dc ). Based on the result from supplementary experiment, Dc can be estimated at the depth separation distance between 4 and 8 cm. Increasing in the depth separation distance beyond Dc does not show an influence on the reduction of the magnitude of the local effect. From the relationship, we can divide the local effect into two components. The first component of the local effect (L1 ) indicated by the black arrow, shows a depth dependent characteristic. The local effect is a decremental function of the depth separation; the greater depth separation, the lower

386



L

∆L1

∆L1

L

Dc

D

∆L2 Dc

D L

∆L2 D

Fig. 8. Relationship between the magnitude of the local effect (L) and the depth separation distance (D). Dc : critical depth separation distance; L1 and L2 : the amount of local effect which is depth dependent and depth independent, respectively.

magnitude of local effect. The second component of local effect (L2 ) indicated by the open arrow shows a depth independent characteristic. Variation of the depth separation distance does not show an influence on this portion of local effect. This means it is not possible to estimate the required depth separation distance where the local effect is vanished. Therefore, we conclude that the depth separation alone can not completely exclude the local effect because a portion of local effect is not depth dependent. Finally, the existence of local effect in the assessment of recognized illuminant has been presented in several works. However, it seems nobody has pay attention to the method for excluding the local effect and getting the valid result. Our work is an attempt to develop a practical method for assessing the valid recognized illuminant. We think this recognized illuminant is an important factor in our new concept for solving color inconsistency among different environment. Generally, color appearance model (CAM) is used to solve the color inconsistency problem. By inputting required data and suitable parameter, CAM can predict color appearance of the test in the targeted environment. However, the disadvantage of CAM is that the accuracy of the color predicted by CAM depended on selecting of the parameter. If the parameter is not appropriate, the color prediction would be failed. Moreover, some parameters related to depth and three-dimensionality are not included in the model. We therefore, proposed another concept called environmentdependent color management system to solve the color inconsistency problem. Instead of selecting the factors to match the target environment, it had better to directly measure the factors from the target environment. We called this process as environment characterization. Once the environment was characterized, we would understand the effect of that environment on the color appearance. It is also helpful to divide the amount of the effect into two components. One is the effect of the entire environment and the other is the local effect. The recognized illuminant in our experiment is supposed to be an important factor representing the environment characteristic. Therefore, it is

necessary to develop a method for assessing the recognize illuminant without involving of the local effect. Even though our result obtained by employing the depth separation in the perceptual achromatic setting seems to be partially successful. There might be another method to exclude the local effect based on the concept of the destruction of belongingness. For example, the translational moving test patch is used instead of the static test patch in the perceptual achromatic setting procedure. Another idea for obtaining the valid color of the recognized illuminant is to conduct the perceptual achromatic setting at several places in that environment. The average result possibly represent the valid color of illuminant because the local effect might be cancelled each other. There might be other claims that under high saturated color illuminant condition, the perceptual achromatic setting cannot be capable for assessing the color of recognized illuminant. The perceptual achromatic locus will not coincide with the chromaticity of physical illuminant because of the incomplete color constancy of our visual system. On the contrary, we did not think the incomplete color constancy is a problem in perceptual achromatic setting. As described in §1, it is not the physical illuminant but the recognized illuminant which plays an important role on the perceived color. And the recognized illuminant does not always coincide with the physical illuminant. This tendency is more common under the colored illuminant than the neutral illuminant such as D65. Since the perceptual achromaticity is supposed to represent the color of the recognized illuminant, the perceptual achromatic setting is still, or even more, important in the incomplete color constancy. Moreover, the local effect (or chromatic induction) seems to be relatively small under the saturated chromatic illuminant. Previous work has shown that the background surface insignificantly bias perceptual achromatic locus.3) Therefore, it can be said that the perceptual achromatic setting is more effective to measure the recognized illuminant under the high saturated color illuminant. 4.

Conclusions

Exclusion of the local effect from the perceptual achromatic setting by depth separation method was partially successful. Even though the depth separation increased, a portion of the local effect still exists. Another method is required to completely exclude the remaining local effect from the assessment of the recognized illuminant. References 1) P. Petrov, C. Y. Kim, I. S. Kweon and Y. S. Seo: Col. Res. Appl. 23 (1998) 159. 2) J. A. Schirillo and S. K. Shevell: Perception 26 (1997) 507. 3) D. H. Brainard: J. Opt. Soc. Am. A 15 (1998) 307. 4) M. Ikeda, H. Shinoda and Y. Mizokami: Opt. Rev. 5 (1998) 200. 5) R. O. Brown and D. I. A. MacLeod: Curr. Biol. 7 (1997) 844. 6) I. Kuriki and K. Uchikawa: J. Opt. Soc. Am. A 15 (1998) 2263. 7) P. B. Delahunt and D. H. Brainard: J. Vision 4 (2004) 57.

OPTICAL REVIEW Vol. 13, No. 5 (2006) 8) J. M. Kraft and D. H. Brainard: Proc. Natl. Acad. Sci. U.S.A. 96 (1999) 307. 9) J. M. Kraft, S. I. Maloney and D. H. Brainard: Perception 31 (2002) 247. 10) A. L. Gilchrist: Science 195 (1977) 185. 11) D. H. Brainard: Spatial Vis. 10 (1997) 433. 12) P. Cunthasaksiri, H. Shinoda and M. Ikeda: Col. Res. Appl. 29 (2004) 255. 13) S. T. Bergstom, G. Derefeldt and S. Holmgren: Scand. J. Psychol. 19 (1978) 265. 14) P. Cunthasaksiri, H. Shinoda and M. Ikeda: Col. Res. Appl. 31


387

(2006) 184. 15) P. Cunthasaksiri and H. Shinoda: submitted to Opt. Rev. 16) H. Aoki, H. Shinoda and M. Ikeda: Proc. 9th Congress Int. Colour Assoc., 2002, p. 307. 17) H. Aoki, H. Shinoda and M. Ikeda: KOGAKU (Jpn. J. Opt.) 32 (2003) 551 [in Japanese]. 18) S. K. Shevell and P. R. Millner: Vis. Res. 36 (1996) 949. 19) J. A. Schirillo and S. K. Shevell: J. Opt. Soc. Am. A 10 (1993) 2442. 20) S. K. Shevell and J. Wei: Vis. Res. 40 (2000) 3173. 21) K. Yamauchi and K. Uchikawa: J. Vision 5 (2005) 515.