Visual Color Matching under Various Viewing Conditions

Transcription

1 Visual Color Matching under Various Viewing Conditions Hitoshi Komatsubara, 1 * Shinji Kobayashi, 1 Nobuyuki Nasuno, 1 Yasushi Nakajima, 2 Shuichi Kumada 2 1 Japan Color Research Institute, Ueno Iwatsuki-shi, Saitama, Japan 2 Canon Inc., Shimomaruko Otaku, Tokyo, Japan Received 2 March 2001; accepted 6 February 2002 Abstract: In this article, we provide colorimetry data for which it was judged that the colors between different media matched under various viewing conditions. Painted color patches, a monitor, and printed color patches were used in the color matching experiments, in which we compared the appearances of the painted color patch and the monitor, or the monitor and the printed color patch, using the method of constant stimuli. The nine types of viewing conditions we used could be envisaged to occur when comparing different device outputs indoors. The experimental data obtained were compared with corresponding colors predicted with the use of five types of color appearance model, including color appearance formulae. We found that when the viewing conditions were the same for the different media, there was good agreement between the experimental data and the CIECAT94 model. And by adjusting the brightness induction and the chromatic induction factors, it was possible to improve conformity for the lightness and the chromaticity. Moreover, it was possible to improve the white point shift, which cannot be adjusted with the use of optimized parameters by introducing incomplete adaptation. By optimizing the parameters and introducing incomplete adaptation, it is possible to make the mean color difference E* ab between the corresponding color and the color matching point less than 10 CIELAB units for all of the viewing conditions Wiley Periodicals, Inc. Col Res Appl, 27, , 2002; Published online in Wiley InterScience ( DOI /col * Correspondence to: Hitoshi Komatsubara, Japan Color Research Institute, Ueno Iwatsuki-shi, Saitama, Japan (Jcri0001@cb.mbn.or.jp) 2002 Wiley Periodicals, Inc. Key words: cross-media; color appearance; color matching experiment; color appearance model; parameter INTRODUCTION Metamerism occurs as a result of the differences in the coloring materials and the color production methods, with the surface colors reproduced by paints or dyes and photographs or printer outputs. In the same way, metamerism occurs between surface colors and monitor outputs. Moreover, monitors make use of the luminescence of phosphors, and so produce a different color mode to that of surface colors. It is known that whether or not the colors produced by such different media match one another depends on the viewing conditions. Braun and Fairchild 1 have reported that when the viewing conditions are the same for the different media, the appearances of the colors on the different media can be matched by colorimetric color matching the colorimetry values. Luo et al. 2 4 have reported data obtained by evaluating the color appearances for surface and luminous color modes under a variety of viewing conditions with the use of a magnitude estimation technique. Mori et al. 5 have reported observation data for the corresponding colors under various chromatic adaptations. However, these reports involved different evaluation techniques and viewing conditions, so whether or not the data are compatible with each other is a problem. The guidelines 6 for coordinated research on evaluation of color appearance between reflection print and self-luminous displays were thus published with CIETC1-27, and it was proposed that observation data be collected. Three evaluation methods successive haploscopic viewing configuration, successive binocular viewing arrangement, and magnitude estimation were proposed. Moreover, Braun et al. 7 have carried out comparative stud- Volume 27, Number 6, December

2 TABLE I. Viewing condition investigation. Exp. no. Print white point (K) CRT white point (K) Background (print/crt) Surround (print/crt) Luminance (cd/m 2 ) (print/crt) Monitor shade S 6,500 6,500 N5Gray/N5Gray Dark/Dark 96/110 N 1 6,500 6,500 N5Gray/N5Gray Light/Light 157/90 N 2 6,500 9,300 N5Gray/N5Gray Light/Light 157/90 N 3 6,500 5,000 N5Gray/N5Gray Light/Light 157/90 N 4 4,200(F6) 6,500 N5Gray/N5Gray Light/Light 157/90 N 5 4,200(F6) 9,300 N5Gray/N5Gray Light/Light 157/90 N 6 5,000(F10) 6,500 N5Gray/N5Gray Light/Light 157/90 N 7 6,500 6,500 N5Gray/N5Gray Light/Light 76/90 N 8 6,500 6,500 N5Gray/N5Gray Light/Light 157/90 Y ies of a number of cross-media image-viewing techniques. In our current research, a painted color patch and a monitor, or a monitor and a printed color patch, were used so as to correspond to simultaneous binocular viewing, and we evaluated the color appearances using the constant method. However, because the color chart and the monitor, or the monitor and the printed color patch, could not both fit into the visual field at the same time, in actual practice, the setup effectively corresponded to a successive binocular viewing arrangment. There are separate ISO stipulations from the International Organization for Standardization (ISO) for the viewing conditions when carrying out visual comparison with color charts 8 and with photographs. 9 However, there are no standard viewing conditions when directly comparing the colors produced by different media. In our current research, we surveyed the trends in ISO deliberations and in Japanese industry, and carried out comparisons under one set of standard viewing conditions and eight sets of practical viewing conditions. We carried out the color appearance comparisons using only color patches, and estimated the color matching points and ranges for which we thought that the colors matched between the painted color patch and the monitor, or the monitor and the printed color patch, using probit analysis, which is a constant method analytic technique. The color matching points were compared with corresponding colors predicted using five types of chromatic adaptation formulae and color appearance models. METHODS Canon Color-Matching Data Viewing Conditions. With the CIE colorimetry system, the following are recommended as standard viewing conditions 10 : Illuminant: Standard illuminant D65 Illuminance: 1,000 lx Background luminance: 64 cd/m 2 Background: Gray (L * 10 50) Sample size: At least 4 These conditions assume evaluation of surface colors in a bright room. However, it is assumed that, in general, a monitor is viewed in a dark room as with the viewing FIG. 1. Layout of experiment. 400 COLOR research and application

3 FIG. 2. Stimulus configuration. conditions for a TV monitor. Moreover, the maximum luminance for a monitor depends on the luminance of the luminescent red, green, and blue phosphors (RGB). When carrying out comparisons between different media such as a color patch and a monitor, it is thus necessary to set conditions that can be realized with both of the media. There are limitations on the luminance range that can be reproduced by a monitor, so viewing conditions that can be reproduced with a monitor but also come as close as possible to satisfying the International Commission on Illumination (CIE) colorimetry system viewing conditions were our standard viewing conditions. Moreover, we presumed that light sources having a whole variety of characteristics are used in industry, so conditions that are close to the actual environments were taken as practical viewing conditions. The standard viewing condition (S) and the practical viewing conditions (sets 1 8) are shown in Table I. Experimental Apparatus. The apparatus used in the experiments and its layout are shown in Figs. 1 and 2. A sample stand with a center height of 275 mm was installed in a light booth equipped with a D65 simulator placed on a table (which was 75 cm above the floor) for the standard condition and directly on a table having a gray background of luminance factor 0.2 for practical conditions. The light booth with a gray background of luminance factor 0.2 was used only in the comparisons carried out under the standard viewing conditions. With the practical viewing conditions, we used the illuminant of color patch on the sample stand with the ambient lighting in the laboratory. A cathode ray tube monitor (CRT) was also placed in the light booth or on the table, with the same center height as that of the center of the sample stand. With mm color patches on a TABLE II. Colorimetric data for the painted color patches. No. Color X Y Z L* a* b* 2 5R5/ R5/ Y5/ G5/ G5/ B5/ B5/ PB5/ PB5/ P5/ P5/ RP5/ RP5/ R6/ YR6/ YR6/ Y6/ GY6/ GY6/ GY6/ G6/ BG6/ B6/ PB6/ P6/ RP6/ R7/ YR7/ Y7/ Y7/ GY7/ GY7/ G7/ BG7/ B7/ PB7/ P7/ Volume 27, Number 6, December

4 FIG. 3. Frequency probability mm gray background of luminance factor 0.2, with a 10mm white border placed on the sample stand and presented on the CRT, we made comparisons for each condition. The observer alternately viewed the color patch on the sample stand and that on the CRT from a distance of 1 m, and judged whether or not the colors matched, with the distance to the center of the sample stand and that of the CRT being 60 cm. The sizes of the color patches corresponded to an angle of view of about 2. The white borders were set as the reference white of the sample stand and the CRT. With a Canon CLC700 color laser printer, we used painted color patches and printed color patches presented on the sample stand. The colorimetric data of the painted color patches by Minolta spectrophotometer CM2002 are shown in Table II. These color patches could be reproduced using the CRT and the printer. The color patches presented on the CRT were displayed on a Barco Reference Calibrater CCID120T driven by a Cambridge Research System visual stimulus generator (VSG). However, the color matching points for the color patches and the CRT color patches vary according to the viewing conditions, so we carried out preliminary experiments using the adjustment method in order to obtain assumed color matching points. The color patches presented on the CRT were produced with color difference in six directions ( hue, hue, lighteness, lighteness, chroma, and chroma) from the predetermined color matching point. And we used an XYZ transformation matrix obtained from the CRT input signal/ luminance characteristic in the color prediction for the CRT TABLE III. Color matching points between the painted color patch and the CRT color patches. No. Color Color matching point Acceptable range L* a* b* H* ab L* ab C* ab 2 5R5/ R5/ Y5/ G5/ G5/ B5/ B5/ PB5/ PB5/ P5/ P5/ RP5/ RP5/ R6/ YR6/ YR6/ Y6/ GY6/ GY6/ GY6/ G6/ BG6/ B6/ PB6/ P6/ RP6/ R7/ YR7/ Y7/ Y7/ GY7/ GY7/ G7/ BG7/ B7/ PB7/ P7/ COLOR research and application

5 TABLE IV. Color matching points between the CRT color patch and the printed color patches. No. Color Color matching point Acceptable range L* a* b* H* ab L* ab C* ab 2 5R5/ R5/ Y5/ G5/ G5/ B5/ B5/ PB5/ PB5/ P5/ P5/ RP5/ RP5/ R6/ YR6/ YR6/ Y6/ GY6/ GY6/ GY6/ G6/ BG6/ B6/ PB6/ P6/ RP6/ R7/ YR7/ Y7/ Y7/ GY7/ GY7/ G7/ BG7/ B7/ PB7/ P7/ color patches. As with the CRT color patches, the printed color patches were produced by determining assumed color matching points between the printed color patches and the CRT color patches through results of preliminary experiments. The printed color patches were also produced with color difference in six directions ( hue, hue, lightness, lightness, chroma, and chroma) from the predetermined color matching points. FIG. 4. Color matching points and acceptable threshold. The circles indicate a* and b* chromaticity of painted color patches; solid lines indicate tolerance vectors of CRT color matching point; dashed lines indicate tolerance vectors of printer color matching point (upper: L* 50; middle: L* 60; lower: L* 70). Volume 27, Number 6, December

6 FIG. 5. Lightness difference of matching points from painted color patches and acceptable threshold. The broken line indicates mean value each lightness level (left: L* 60; middle: L* 70; right: L* 80). The CIELAB values for the painted, the CRT, and the the printed color patches were measured under each of the sets of viewing conditions with a Topcon BM-7 luminance colorimeter. Psychological Evaluation Method. When comparing a surface color with the CRT color produced by inputting this surface color with an input device such as scanner or digital camera, we perceived a color difference even under the standard viewing conditions. We thought that it should be possible to achieve colorimetric color matching under the standard viewing conditions. However, it is not possible to reduce the color difference to zero, due to factors such as the profile accuracies of the input device and the CRT. There is thus a problem as to how large the color difference between media may be before it becomes unacceptable. In our current research, color patches were thus compared with one another visually, and the color matching point and acceptable range were determined. The following example illustrates the experimental procedure comparing painted color patches with CRT color patches. Observers view the gray background in the light booth in order to adapt to the illuminant. One of the color patches shown in Table II is selected at random and placed on the sample stand as a reference sample. Observers first view the reference sample, and then the color patches presented at random, one after another on the CRT, with each either corresponding to the predetermined assumed color matching point or else having hue, lightness, and chroma that differ from the assumed color matching point. Observers successively compare the painted color patch with the CRT color patch by shifting their line of sight, and evaluating the color difference between the the reference color and the CRT color patch in accordance with the following judgment criteria: 1. Color difference is acceptable. 2. Cannot say whether color difference is acceptable or unacceptable. 3. Color difference is unacceptable. After completing their evaluation for a color patch, observers input their judgment results into the VSG using number keys; then, the next color patch is presented. Once judgment has been completed for all the prepared CRT color patches, the reference color patch on the sample stand is changed. Under the standard viewing condition, each observer compared the reference painted color patch with CRT color patches, and the reference CRT color patch having similar color matching points between a painted color patch and a CRT color patch with printed color patches. Under the practical viewing conditions, each observer only compared the reference CRT color patch with printed color patches by TABLE V. Average color difference of color matching point from painted color patch, and average acceptable ranges. Average color difference Value CRT color patch Printed color patch H* ab L* ab C* ab E* ab H* ab L* ab C* ab E* ab Average Average acceptable range CRT color patch printed color patch value H* ab L* ab C* ab E* ab H* ab L* ab C* ab E* ab Average COLOR research and application

7 TABLE VI. Color matching points and acceptable ranges between the printed color patch and the CRT color patch for cases 1 and 2. Case 1 Case 2 Color No. Printed color patch Color matching point Acceptable range Color matching point Acceptable range L* a* b* L* a* b* H* ab L* ab C* ab L* a* b* H* ab L* ab C* ab the same experimental procedure as that used for comparing painted color patches with CRT color patches. Each observer carried out three evaluations for the color patches in Table I. The 4 observers for the standard condition and 7 observers for other conditions all had normal color perception. RESULTS Evaluation Results under Standard Viewing Conditions Figure 3 shows a graph of the average probability of observers for color patch No. 8 under the standard viewing conditions against the color difference from the assumed color matching point. This was calculated by taking the probability for judgment criterion 1 to be 1, that for criterion 2 to be 0.5, and that for criterion 3 to be 0, thus corresponding to the probability of acceptance. Because the distribution of the probability is approximated by an ogive, the color corresponding to a probability of 50% can be taken as being the acceptable threshold for the color patch in question. In actual practice, the threshold can be caluculated by using the probit analysis, 11 which is a method of predicting the threshold for method of constant stimuli. The center of Fig. 3 is the same as the assumed color matching point, so it is not necessarily the case that the probability distribution reaches its maximum for this color. In some cases, the results may lean either to the positive or the negative side. In such cases, the threshold on the positive or negative side was estimated after the assumed color matching point was shifted to the color for which the probability was estimated as maximum. The results of estimating the threshold between the reference color patch and the CRT color patches are shown in Table III. The color matching point in Table III indicates the midpoint of the positive- and negative-side acceptable threshold. This is because (as shown in Fig. 3) the probability distribution curves on the positive and negative sides have gently sloping figures, so the midpoint was Volume 27, Number 6, December

8 TABLE VII. Color matching points and acceptable ranges between the printed color patch and the CRT color patch for cases 3 and 8. Case 3 Case 8 Color No. Printed color patch Color matching point Acceptable range Color matching point Acceptable range L* a* b* L* a* b* H* ab L* ab C* ab L* a* b* H* ab L* ab C* ab taken as the color matching point. The acceptable range thus indicates the difference between the positive and negativeside threshold and the color matching point in Table III. Table IV shows the color matching points between the printed color patches and the reference CRT color patch, along with the acceptable ranges. However, the colorimetric value of the reference CRT color patch used in the comparison does not equal the colorimetric value of the color matching point between the painted color patch and CRT color patches in Table II; therefore,the color matching point was calculated by performing the following correction, which was done in order to compare the painted/crt color matching point with the CRT /printed color matching point. Here, X (X CRT,CMP /X CRT,MEA ) X Y (Y CRT,CMP /Y CRT,MEA ) Y Z (Z CRT,CMP /Z] CRT,MEA ) Z (1) X, Y, Z : Color matching point after correction X, Y, Z: Color matching point before correction X CRT,MEA, Y CRT,MEA, Z CRT,MEA : Measurement value for the reference CRT color patch X CRT,CMP, Y CRT,CMP, Z CRT,CMP : Color matching point for the painted color patch and the CRT color patch in Table II. Figures 4 and 5 show the relationship between the colorimetric values of the painted color patches, the color matching points, and the acceptable ranges. Moreover, Table V also shows the average color difference between color matching point and painted color patch, and average acceptable threshold.the hue, lightness, and chroma differences for CRT color patches and printed color patches are less than 3 CIELAB units, so it appears that colorimetric color reproduction is more or less achieved. However, with regard to the lightness, there is a tendency for the colors to match when the value is higher for the printed color patch than that of the CRT color patch, and there is a significant difference between the color matching points for the CRT color patch and printed color patch. It is thought that this tendency is due to the effects of the glossiness of the printed color patch surface on the colorimetry and the color appearance. These effects are also seen slightly with the chromaticity color matching points, with the color matching points for the printed color patches being shifted slightly to the high chroma side. However, unlike the case with the lightness, 406 COLOR research and application

9 TABLE VIII. Color matching points and acceptable ranges between the printed color patch and the CRT color patch for cases 4 and 5. Case 4 Case 5 Color No. Printed color patch Color matching point Acceptable range Color matching point Acceptable range L* a* b* L* a* b* H* ab L* ab C* ab L* a* b* H* ab L* ab C* ab the chromaticity of the color matching points for the CRT color patches and the printed color patches are both more or less in acceptable ranges and there is no significant difference. Evaluation Results under Practical Viewing Conditions Tables VI through X show, for the eight sets of practical viewing conditions, color matching points between the CRT color patches and the printed color patches, along with the acceptable threshold. The measurement values for the printed color patches were measured with a color luminance meter for each of the sets of viewing conditions. The chromaticity color matching points for Tables VI through X are shown in Figures 6 through 9. Figure 6 shows the results for cases 1 3, and Figure 7, the results for cases 4 and 5. Also, Figure 8 shows the results for case 6, and Figure 9, the results for cases 7 and 8. The difference between cases 1 and 8 lies in whether or not the CRT was fitted with a shading hood, and the results for the two agree well with one another. With the exception of low lightness (L* 60), colorimetric color reproduction is achieved for both cases. With low lightness (L* 60), for the colors that show a tendency for color matching to be on the high chroma side, it is thought that the reduction in the luminance contrast due to the stimulus and the background being set to the same luminance had an effect. Moreover, comparing case 7 with case 0, case 7 was carried out not in a dark room but rather in a bright room with the same light as ambient light, and the illuminance of the printed color patches was lowered. As a result, the lightness color matching point for the CRT color patches tended to be lower. The chromaticity of color matching point, on the other hand, showed virtually no difference at high lightness (L* 80), tending to be greater when the lightness was lower. These trends are due to the fact that the reflected glare with the ambient lighting reduces the brightness contrast between stimulus and background, and that the apparent brightness of the CRT color patches was raised due to the low illuminance of the printed color patches. With cases 2 6, it is clear that colorimetric color reproduction is not realized, showing that it is necessary to carry out studies into the corresponding color with the use of chromatic adaptation formulae or color appearance models. Corresponding Colors from Chromatic Adaptation Formulae and Color Appearance Models It is clear from the evaluation results under the practical viewing conditions that there are large discrepancies from Volume 27, Number 6, December

10 TABLE IX. Color matching points and acceptable ranges between the printed color patch and the CRT color patch for case 6. Case 6 Color No. Printed color patch Color matching point Acceptable range L* a* b* L* a* b* H* ab L* ab C* ab colorimetric color reproduction. We thus predicted the printed color patch and CRT color patch color matching points using chromatic adaptation formulae and color appearance models, and then compared the prediction and evaluation results. Six types of chromatic adaptation formulae and color appearance model were used CIECAT94, 12 CMCCAT97, 13 CIECAM97s, 14 HUNT94, 15,16 NAYA- TANI, 17,18 and RLAB. 19 For each of the models, the recommended standard values were used for the environmental parameters (chromatic, brightness, luminance level induction factors, etc.). Note, however, that with regard to the incomplete adaptation proposed by Nayatani et al., we assumed when carrying out the calculations that there is incomplete adaptation for all of the models. Tables XI through XIII show the average color differences between the corresponding colors predicted with use of the 6 types of model and the color matching points from the evaluation experiments, along with the standard deviations. Specifically, Table XI summarizes the results for all of the samples for cases 1-8. Tables XII and XIII show the average color difference and standard deviation for each lightness of color patch. As an example of the difference between the corresponding color and the color matching point, the results for case 2 and high lightness (L* 80) are shown in Figure 10. Regarding the color differences between the corresponding colors that we calculated using each of the color appearance models and the color matching points, the predicted values are bright for all of the models and the order is CMC- CAT97, RLAB NAYATANI CIECAT94 CIECAM97 HUNT94, showing that CMCCAT97 and RLAB are good. And the order is CIECAT94 CIECAM97 CMCCAT97 HUNT94 RLAB, NAYATANI for the chromaticity. For the overall color difference, the order is CIECAT94 CMCCAT97 RLAB NAYATANI, CIECAM97 HUNT94, showing that the conformity is good. Looking at the results for each lightness of color patch, it can be seen that the lighter the sample, the greater the tendency toward good conformity for all of the models. There is no marked difference in the chromaticity according to the lightness of color patch, and it 408 COLOR research and application

11 TABLE X. Color matching points and acceptable ranges between the printed color patch and the CRT color patch for case 7. Case 7 Color No. Printed color patch Color matching point Acceptable range L* a* b* L* a* b* H* ab L* ab C* ab is presumed that this is because the change in color appearance accompanied by increase in the luminance contrast between the CRT color patches and the background affected the judgment of lightness matching. Figure 10 shows a representative example of the difference between the corresponding color and the color matching point in case 2; it shows that for all of the models, the corresponding color tends to be shifted in the chromaticity direction of the illuminance relative to the color matching point. DISCUSSION Parameter Effects in Chromatic Adaptation Formulae and Color Appearance Models It is known that the agreement of colors between different media depends on the viewing conditions. 1 We have already discussed the fact that the tristimulus values are different with regard to color matching points between the printed colors and the monitor colors carried out under eight types of lighting environment, and that there is a clear discrepancy from colorimetric color reproduction. These lighting environments were selected by imagining the situations in which colored images would be evaluated in an office space. The results of comparing color matching points with corresponding colors predicted by use of the recommended values of parameters for each model indicated that there is a tendency for the brightness and chromatic inductions of the background to be overestimated, and that there is a white point mismatch. Recommended values of color appearance model parameters such as the brightness and chromatic induction factors of the background have been stipulated separately for each of a number of typical viewing condition categories. And when the stipulated parameter values are used, 20 it has been reported that the color appearance models do not necessarily give good predictions depending on Volume 27, Number 6, December

12 FIG. 6. Chromaticity and lightness of color matching point and acceptable range. Arrow indicates the shift of color matching point from printed color patch and arrowhead indicates the color matching points. The axis of ellipses is defined by acceptable range: (a) shows the results for case 1 (left: L* 60; middle: L* 70; right: L* 80); (b) shows the results for case 2; and (c) shows the results for case COLOR research and application

13 FIG. 7. Chromaticity and lightness of color matching point and acceptable range. Arrow indicates the shift of color matching point from printed color patch, and arrowhead indicates the color matching points. The axis of ellipses is defined by acceptable range: (a) shows the results for case 4 (left: L* 60; middle: L* 70; right: L* 80); (b) shows the results for case 5. FIG. 8. Chromaticity and lightness of color matching point and acceptable range for case 6. Arrow indicates the shift of color matching point from printed color patch, and arrowhead indicates the color matching points. The axis of ellipses is defined by acceptable range (left: L* 60; middle: L* 70; right: L* 80). Volume 27, Number 6, December

14 FIG. 9. Chromaticity and lightness of color matching point and acceptable range. Arrow indicates the shift of color matching point from printed color patch, and arrowhead indicates the color matching points. The axis of ellipses is defined by acceptable range: (a) shows the results for case 7 (left: L* 60; middle: L* 70; right: L* 80); (b) shows the results for case 8. TABLE XI. Average color difference between the corresponding color and the color matching point, and average standard deviation of color difference. CIECAT94 CMCCAT97 CIECAM97 HUNT94 NAYATANI RLAB L E(ab) E L E(ab) E L E(ab) E L E(ab) E L E(ab) E L E(ab) E Case 1 Average SD Case 2 Average SD Case 3 Average SD Case 4 Average SD Case 5 Average SD Case 6 Average SD Case 7 Average SD Case 8 Average SD Total Average SD COLOR research and application

15 TABLE XII. Average color difference between the corresponding color and the color matching point for each lightness level. CIECAT94 CMCCAT97 CIECAM97 HUNT94 NAYATANI RLAB Lightness level L E(ab) E L E(ab) E L E(ab) E L E(ab) E L E(ab) E L E(ab) E V 5 Case Case Case Case Case Case Case Case Total V 6 Case Case Case Case Case Case Case Case Total V 7 Case Case Case Case Case Case Case Case Total the viewing method and the viewed images. Moreover, Nayatani et al. have pointed out that it is necessary to give consideration to incomplete adaptation to the white point of the test light, and have proposed introducing an effective adapting coefficient into the chromatic adaptation formula. 21 TABLE XIII. Average standard deviation of color differences between the corresponding color and the color matching point. CIECAT94 CMCCAT97 CIECAM97 HUNT94 NAYATANI RLAB Lightness level L E(ab) E L E(ab) E L E(ab) E L E(ab) E L E(ab) E L E(ab) E V 5 Case Case Case Case Case Case Case Case Total V 6 Case Case Case Case Case Case Case Case Total V 7 Case Case Case Case Case Case Case Case Total Volume 27, Number 6, December

16 TABLE XIV. Types of parameter and simulation ranges. Model Brightness induction factor of background (c or Nb) Parameter Chromatic induction factor of background (Nc) Effective adapting coefficient CIECAT CIECAM97s HUNT NAYATANI FIG. 10. Result of color difference between color matching points and corresponding colors for L* 80. Arrow indicates the shift of corresponding color from printed color patch, and arrowhead indicates the chromaticity of corresponding color. In our current research, we carried out studies into optimal values for the brightness induction factor, the chromatic induction factor, and the effective adapting coefficient by using the results of visual color matching between printed color patches and monitor colors under the eight types of lighting environment shown in Table I. In addition, we compared the performances of five types of color appearance model CIECAT94, CMCCAT97, CIECAM97, HUNT94, and NAYATANI based on the corresponding colors predicted using the optimized parameters. However, for the CIECAT94, CMCCAT97, and NAYATANI models, the effective adapting coefficient was the only parameter. Parameters The parameters used in the various color appearance models are shown in Table XIV, along with the values of these TABLE XV. Parameter effect about brightness induction factor and chromatic induction factor of background for CIECAM97s. CIECAM case 1 CIECAM case 2 CIECAM case 3 CIECAM case 4 Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) c L E(ab) E(Lab) c L E(ab) E(Lab) c L E(ab) E(Lab) c L E(ab) E(Lab) CIECAM case 5 CIECAM case 6 CIECAM case 7 CIECAM case 8 Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) c L E(ab) E(Lab) c L E(ab) E(Lab) c L E(ab) E(Lab) c L E(ab) E(Lab) COLOR research and application

17 TABLE XVI. Parameter effect about brightness induction factor and chromatic induction factor of the background for HUNT94. Hunt case 1 Hunt case 2 Hunt case 3 Hunt case 4 Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nb L E(ab) E(Lab) Nb L E(ab) E(Lab) Nb L E(ab) E(Lab) Nb L E(ab) E(Lab) Hunt case 5 Hunt case 6 Hunt case 7 Hunt case 8 Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nc L E(ab) E(Lab) Nb L E(ab) E(Lab) Nb L E(ab) E(Lab) Nb L E(ab) E(Lab) Nb L E(ab) E(Lab) parameters used in the predictions. The recommended values for the parameters under the viewing conditions shown in Table XIV are as follows: CIECAT94: 1; CMCCAT97: 1; CIECAM97: c 0.69, Nc 1, 1; HUNT94: Nb 75, Nc 1, 1; NAYATANI: 1. Brightness Induction Factor of the Background We predicted the corresponding colors for the monitor colors by using the tristimulus values of the printed color patch for which color matching had been carried out under the viewing conditions shown in Table I. We found the parameters for which the color difference (CIELAB) between the color matching point of printed color and the corresponding color is minimal. However, when the values of the brightness or chromatic induction factor were changed separately, broadly speaking, predicting the corresponding color only changed the brightness or the chromaticity. Therefore, we can assume a parameter effect of the FIG. 11. Lightness difference between the color matching point and the corresponding color against various brightness induction factors about CIECAM97s. FIG. 12. Lightness difference between the color matching point and the corresponding color against various brightness induction factors about HUNT94. Volume 27, Number 6, December

18 FIG. 13. Chromaticity difference between the color matching point and the corresponding color against various chromatic induction factors about CIECAM97s. FIG. 14. Chromaticity difference between the color matching point and the corresponding color against various chromatic induction factors about HUNT94. TABLE XVII. Parameter effect about adapting coefficients for CIECAT94. CIECAT case 1 CIECAT case 2 CIECAT case 3 CIECAT case 4 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) CIECAT case 5 CIECAT case 6 CIECAT case 7 CIECAT case 8 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) TABLE XVIII. Parameter effect about adapting coefficients for CAMCAT97. CAMCAT97 case 1 CAMCAT97 case 2 CAMCAT97 case 3 CAMCAT97 case 4 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) CAMCAT97 case 5 CAMCAT97 case 6 CAMCAT97 case 7 CAMCAT97 case 8 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) brightness induction factor that is independent of other factors. And so this study was carried out in the corresponding color only in the case of a changing brightness induction factor, because the chromatic induction factor and the effective adapting coefficient were fixed at their recommended values. The results for the CIECAM97s and HUNT94 models are shown in Tables XV and XVI. Figures 11 and 12 show graphs of the lightness difference ( L*) between the color matching point of prined color and the corresponding color against the brightness induction factor. For both the CIECAM97s and HUNT94 models, we found a correlation between the parameter value and the lightness 416 COLOR research and application

19 TABLE XIX. Parameter effect about adapting coefficients for CIECAM97s. CIECAM97 case 1 CIECAM97 case 2 CIECAM97 case 3 CIECAM97 case 4 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) CIECAM97 case 5 CIECAM97 case 6 CIECAM97 case 7 CIECAM97 case 8 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) difference ( L*), with the optimal values being c for the CIECAM97s and Nb for the HUNT94. These optimal values are lower than the recommended values for the models and are close to the recommended values for a dim surround. In this case, we performed color matching experiments under bright lighting enviroments and with monitor colors having a three-part structure consisting of a white border, a background, and a test color that looked like the surface color mode. Therefore, we presumed that the color appearance of monitor colors depends on the stipulations. There were virtually no differences in the optimal values between the different sets of viewing conditions, although the optimal values for cases 4 and 5 were lower than those for the other cases. With cases 4 and 5, the difference in color temperature between the illuminant for printed color and the monitor white was large, so we presumed that there were effects due to the visual clarity received from the white on the prints (4,200 K). This suggests that it is necessary to consider the effects of incomplete adaptation on the brightness induction factor. The mean optimal values over the eight sets of viewing conditions are c for CIECAM97s and Nb 42 for HUNT94. Chromatic Induction Factor of the Background Similar to the procedure used with the brightness induction factor of the background, only the chromatic induction factor of the background was changed. Therefore, the brightness induction factor of the background and the effective adapting coefficient were fixed at their recommended values. The color difference between the color matching point of printed color and the corresponding color for this case is shown in Tables XV and XVI. And the relationship between the chromatic induction factor and the chromaticity difference ( E* ab ) between the color matching point and the corresponding color is shown in Figures 13 and 14. With the CIECAM97s model, the results vary for different sets of viewing conditions, and the optimal value for cases 4 and 5 tend to be lower than for the other cases. For other cases, the chromatic induction factor for which the chromaticity difference is a minimum lies in the range Nc , and the effect of the parameter is larger than that in cases 4 and 5. As with the brightness induction factor, the optimal value of the chromatic induction factor is close to the value for a dim or dark surround, showing that chromatic induction is acting excessively. With cases 4 and 5, TABLE XX. Parameter effect about adapting coefficients for HUNT94. Hunt case 1 Hunt case 2 Hunt case 3 Hunt case 4 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) Hunt case 5 Hunt case 6 Hunt case 7 Hunt case 8 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) Volume 27, Number 6, December

20 TABLE XXI. Parameter effect about adapting coefficients for NAYATANI. NAYATANI case 1 NAYATANI case 2 NAYATANI case 3 NAYATANI case 4 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) NAYATANI case 5 NAYATANI case 6 NAYATANI case 7 NAYATANI case 8 L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) L E(ab) E(Lab) because there are large, incomplete adaptation effects due to the difference in the white point for the prints and the monitor, the effect of the parameter is small. Compared with the brightness induction factor, the effect of the chromatic induction factor is small, because there is no marked change in the chromaticity difference between the case in which the optimal value and that in which the recommended value is used. With the HUNT94, the extent of parameter effect differs according to the viewing conditions, but the trend is largely the same as that obtained with the CIECAM97s. The optimal value for which the chromaticity difference is a minimum lies in the range Nc , which corresponds to a dark surround. However, there is little difference when the recommended value of Nc 1.0 is used. The mean optimal values over the eight sets of viewing conditions are Nc 0.85 for both CIECAM97s and HUNT94. Effective Adapting Coefficient The color difference between the color matching point of printed color and the corresponding color in the case of the effective adapting coefficient is introduced in Tables XVII through XXI. We calculated the corresponding color for CIECAM97s and HUNT94 using the average optimal values over the eight sets of viewing conditions for the brightness and chromatic induction factors of the background. The effective adapting coefficient was used to correct for incomplete adaptation to the chromaticity of the white point of the print surface and the monitor, so it had virtually no effect on the lightness/brightness difference between the color matching point of printed color and the corresponding color. The relationship between the effective adapting coefficient and the chromaticity difference between the color matching point and the corresponding color is shown in Figures 15 through 19. In each of the figures, for cases 1, 7, and 8, the white point is the same for the print surface and the monitor, so there is no change in the chromaticity difference upon changing the effective adapting coefficient. The parameter effect shows the same kind of characteristics for all of the CIECAT94, CMCCAT97, CIECAM97s, and HUNT94, and the chromaticity difference becomes a minimum around However, there tends to be a slight shift to a lower value of for the NAYATANI. The parameter effects of the effective adapting coefficient are marked for cases 4 and 5, in which the print white point is 4,200 K. In these cases, the white point of print surface differs consid- FIG. 15. Chromaticity difference between the color matching point and the corresponding color against various adapting coefficients about CIECAT94. FIG. 16. Chromaticity difference between the color matching point and the corresponding color against various adapting coefficients about CMCCAT COLOR research and application

21 FIG. 17. Chromaticity difference between the color matching point and the corresponding color against various adapting coefficients about CIECAM97s. FIG. 19. Chromaticity difference between the color matching point and the corresponding color against various adapting coefficients about NAYATANI. erably to 5,500 6,500 K, which is the color temperature that Hunt and Winter 22 say is perceived as white, and this shows that adaptation with large color temperature difference is not completely realized. The optimal value 0.7 agrees fairly well with the experimental results of Yano et al. 23 and Kato. 24 With cases 2, 3, and 6, for which the effects of the effective adapting coefficient are small, the optimal value varies somewhat according to the color appearance model. However, even if 0.7 is used, there is no marked change in the color difference between the cases in which the optimal value for respectively viewing conditions is used, so the effective adapting coefficient 0.7 can for practical purposes be used regardless of the viewing conditions. CONCLUSIONS 1. The tristimulus values of the color matching points for painted color patches and CRT color patches placed in a dark room under the same viewing conditions more or less matched. Even with a color chart/crt/printer serial system, with the color patches used as the inputs, more or less the same results were obtained. FIG. 18. Chromaticity difference between the color matching point and the corresponding color against various adapting coefficients about HUNT The results for cases 2 to 6, in which reflective samples were illuminated with the ambient light in a bright room, showed clearly that colorimetric color reproduction is not realized. 3. As a result of comparing color matching points with corresponding colors predicted using chromatic adaptation formulae and color appearance models, it was found that the order from best to worst agreement was CIECAT94 RLAB NAYATANI CIECAM97 HUNT The results for the corresponding colors showed that they are brighter than the color matching points, and there tends to be a shift in the chromaticity direction of the chromaticity coordinate of the illuminant. This suggests that it is necessary to carry out studies on the values of the parameters used in the prediction, such as the brightness induction factor, and on incomplete adaptation. 5. We studied the effects of the parameters of color appearance models by using data from color matching experiments and comparing samples having a three-part structure consisting of a white border, a background, and a test color. The parameters studied were the brightness and chromatic induction factors of the background and the effective adapting coefficient. We found that the effects of the parameters for the CIECAM97s and HUNT94 show more or less the same trends, with the optimal values of the parameters being slightly lower than the recommended values for the models. Moreover, the results indicate that it is possible to predict accurately the observation results by introducing the effects of incomplete adaptation to the white point. When the optimal parameter values are used, the color difference between the color matching point of printed color and the corresponding color averaged over the eight sets of viewing conditions is E* 5.1 for the CIECAM97s (Nc 0.85, c 0.625, and 0.7) and E* 7.4 for the HUNT94 (Nc 0.85, Nb 42, and 0.7). With the CIECAT94, CMCCAT97, and NAYATANI, only the effective adapting coefficient can be used, but we found that the effective adapting coefficient has large effects Volume 27, Number 6, December