Impact of cross-regional differences on color rendition evaluation of white light sources

Kevin A.G. Smet; Peter Hanselaer

doi:10.1364/OE.23.030216

1. Introduction

The concept of memory colors, i.e. the colors associated with familiar objects in long-term memory, was first introduced by Hering in the late 19th century who stated that we view the world through the spectacles of memory [1]. Because of their potential use as an internal reference to assess color appearance, memory colors and preferred colors—respectively the colors we remember or would like objects to have—have been and still are of great interest to many areas of color research, such as color reproduction [2–9] and color rendering [10–13].

For example, in case of the evaluation of the color rendition properties of a light source, which usually involves a comparison of object color appearance under a test source with that under a reference illuminant, one could replace the latter with the colour appearance as stored in memory as long as the test objects are very familiar objects. The assumption is that light sources which render the colors of familiar objects similar to their memory (or preferred) colors will result in high color rendition quality in general. The advantage of this approach is that it doesn’t require any optimal reference illuminant. Indeed, it has been shown that the reference illuminants proposed by the Commission International d’Eclairage (CIE) for the calculation of the CIE color rendering index (CRI) [14]—a blackbody radiator or daylight phase of the same correlated color temperature (CCT) as the test source for CCTs respectively below or above 5000 K—, and which have also been implemented in several other color rendition metrics [15–19], are not optimal with respect to subjective aspects of color rendition such as visual appreciation, preference and naturalness [20].

Despite their potential and obvious advantages, memory colors or preferred colors have so far only been implemented in four color rendition metrics: Sanders’ preferred color index R_p [11], Judd’s flattery index R_f [10], Thornton’s color preference index CPI [13] and Smet’s memory color rendition index R_m [12, 21]. Note that Sanders’ R_p and Smet’s R_m are the only two metrics that have explicitly implemented memory or preferred colors as references, as both Judd’s R_f and Thornon’s CPI use a preferred shift—calculated based on memory and preferred color data obtained from literature—to adjust the chromaticity of a set of Munsell test samples illuminated by the typical CIE reference illuminant. For a detailed analysis and discussion of the predictive performance of these metrics with respect to psychophysical data on visual appreciation and naturalness, the reader is kindly referred to Smet et al. [20].

A key question to answer when using memory colors, or even specific illuminants for that matter, as a reference in any sort of metric that assesses color appearance, is whether the chosen reference are region-dependent or culture-dependent. In a recent study, involving laboratories from seven different geographic regions, the memory colors of a set of eleven familiar objects were investigated [22]. Although, a statistical analysis had shown a significant effect of geographical region, the effect size was found to be of the same order as or smaller than the inter-observer variability within a single geographical region, suggesting the effect is of little or no practical relevance.

The current paper specifically investigates the impact of the observed geographic differences of memory colors of some familiar objects on the evaluation of color rendition. First, seven regional and one global memory color rendition metric are developed, similar to Smet’s R_m, based on the regional color appearance rating data. Second, the impact of geographic region is investigated using two approaches. A first approach examines the impact on the regional memory color index values of a database of 401 light sources, while a second one focuses on the impact on the predictive performance—i.e. correlation with visually obtained color rendition data—of the various regional memory color rendition metrics.

2. Regional Memory Color Rendition Indices, R_m,reg

The following subsections describe the design of the regional memory color rendition indices, R_m,reg. Each regional memory color rendition index is composed of three key components: the similarity functions describing the normalized observer response with respect to the memory color (section 2.1), the spectral reflectance of each of the familiar objects (section 2.2) for the calculation of the objects’ appearance under the light source and its corresponding memory color similarity (section 2.3) and a rescaling function to obtain a more familiar CRI-like scale (section 2.4).

2.1 Similarity functions

In a recent study of Smet et al. [22], the memory colors of eleven familiar objects (Green Apple, Ripe Banana, Ripe Lemon, CAuliflower, ORange, StrawBerry, TOmato, Dried Lavender, SMurf®, Caucasian Skin and Asian Skin) were investigated in seven different geographic regions: BElgium, HUngary, BRazil, COlombia, ChiNa, TaiWan and Iran. At each location, identical experiments were conducted in which each familiar object was presented on a calibrated display in a large number (average 165) of different colors to a panel of test subjects. The subjects were asked to rate the similarity of the presented object color with respect to what they remember the object to look like. The adaptation state of the observers was controlled presenting the object surrounded by a white background (CIE illuminant D65 with Y₁₀ = 200 cd/m²). Worldwide over 210 000 ratings were obtained, on average 30 000 per region and 2700 per region per object. For each geographic region reg (reg = BE, HU, BR, CO, TW, CN and IR) and each object i a region average observer was calculated by taking the arithmetic mean of the observer ratings. The region average observers were then modeled with bivariate Gaussian surfaces:

d_{i, r e g} (u_{10}^{'}, v_{10}^{'}) = {{[(\begin{matrix} u_{10}^{'} \\ v_{10}^{'} \end{matrix}) - (\begin{matrix} a_{3, i, r e g} \\ a_{4, i, r e g} \end{matrix})]}^{T} [\begin{matrix} a_{1, i, r e g} & a_{5, i, r e g} \\ a_{5, i, r e g} & a_{2, i, r e g} \end{matrix}] [(\begin{matrix} u_{10}^{'} \\ v_{10}^{'} \end{matrix}) - (\begin{matrix} a_{3, i, r e g} \\ a_{4, i, r e g} \end{matrix})]}^{\frac{1}{2}}

R_{i, r e g} (u_{10}^{'}, v_{10}^{'}) = a_{7, i, r e g} + a_{6, i, r e g} \cdot e^{- \frac{1}{2} d_{i, r e g} {(u_{10}^{'}, v_{10}^{'})}^{2}}

with $a_{1 - 7}$ fitting parameters; $d (u_{10}^{'}, v_{10}^{'})$ the Mahalanobis distance—which defines an elliptical contour of equal ratings —and $R (u_{10}^{'}, v_{10}^{'})$ the rating at the chromaticity coordinate $(u_{10}^{'}, v_{10}^{'})$ . The subscript ‘10’ denotes the use of the CIE 1964 (10°) observer in the calculation of the chromaticity coordinates. The memory color is determined by the center $(a_{3}, a_{4})$ of the bivariate rating function $R (u_{10}^{'}, v_{10}^{'})$ , while $a_{1}, a_{2} and a_{5}$ determine its size, shape and orientation. The parameters $a_{6} and a_{7}$ determine the scale range of the ratings for each object. As an example, the rating functions for a ‘ripe banana’ for each of the regions is illustrated in Fig. 1.

Fig. 1 Bivariate Gaussian models for the observer rating data of a ripe banana.

Download Full Size | PDF

The bivariate rating functions $R (u_{10}^{'}, v_{10}^{'})$ for a global average observer (reg = GL) were obtained by modeling the pooled region average observers ratings of all geographic regions for each object. Full details and model parameters can be found in [22].

Although the rating functions $R (u_{10}^{'}, v_{10}^{'})$ could be used in a memory color rendition metric, it is however more appropriate to use the same scale for each familiar object to ensure each object contributes equally to the average score. Therefore, similarity functions, $S_{i, r e g} (u_{10}^{'}, v_{10}^{'})$ , were obtained that describe the degree of similarity of objects colors (chromaticity only) with their memory color on a zero-to-one scale by setting the model parameters a₆ and a₇ in Eq. (1)b to respectively one and zero.

2.2. Spectral reflectance

The spectral reflectance of most of the familiar objects were determined from spectral radiance measurements of the objects and of a standard white tile with known spectral reflectance. For each object, several measurements were made and the results were averaged. The spectral reflectance for dried lavender was taken to be the one used in Smet’s memory color rendition index as no dried lavender was available at the time of measurement. The spectral reflectance for the Caucasian and Asian Skin were taken as the average of the spectral reflectance functions of two skin databases, i.e. the OULU [23] and RIT [24] sets. The spectral reflectance of the eleven familiar objects, as adopted in the regional MCRIs, are illustrated in Fig. 2.

Fig. 2 Spectral reflectance of the familiar objects.

Download Full Size | PDF

2.3 Calculation of the general similarity indicator, S_a,reg

A regional MCRI general indicator value, S_a,reg of a light source for region reg—indicative of the source’s overall color rendition quality for the region average observer—is calculated as follows.

First, the tristimulus values of the eleven familiar objects are calculated from their spectral reflectance, the spectral radiance of the light source and the CIE 1964 colour matching functions. These values are than transformed to their corresponding colours under the CIE illuminant D65 (i.e. the adaptation white point during the psychophysical experiments) using the CAT02 chromatic adaptation transform. The degree of adaptation is either calculated from the adapting field luminance or set to 0.9 in case the former is unknown. A value of 0.9 is advised as chromatic adaptation is rarely complete, especially for light sources with a less than neutral appearance. Secondly, the corresponding tristimulus values are transformed to the CIE 1976 u’v’ chromaticity coordinates. Thirdly, for each object i specific similarity indicators S_i,reg are calculated by using the appropriate—i.e. the ones corresponding to the region in question—model parameters. A general similarity indicator, S_a,reg is then obtained by taking the geometric mean of the specific similarity indicator values S_i,reg of all objects.

2.4 Rescaling the S_a,reg indicator to the R_m,reg index

Finally, the general similarity indicator S_a,reg, is transformed to a general memory color rendition index, R_m,reg, with a CIE-CRI-like scale using a specific scaling function.

In the MCRI published by Smet, et al. [12] a psychometric scaling function was used with the parameters chosen to ensure indicator values below 0.5 result in index values of 0 ànd CIE illuminants D65 and F4 have index values of respectively 90 and 50. However, using a similar scaling function whereby the scaling parameters are optimized for each of the geographic regions, would automatically result in nearly region independent index values. Since the main goal of this paper is to investigate the impact of cross-geographical differences on color rendition evaluation, this is not an appropriate approach as the rescaling might have a masking effect. Therefore, a simple linear rescaling has been applied for now, i.e. f(x) = 100 - a*(1-x), whereby the coefficient a has been set to 50. This results in a CIE-CRI-like scale whereby the mean R_m,reg for the global average observer of the CIE illuminants F1-F12 approximately equals their mean CIE R_a value. This is a common rescaling approach for color rendition metrics, as this way metrics are assumed to operate on approximately comparable scales [16]. The global average observer was chosen as reference, to specifically investigate the possibility of proposing a single memory color rendition metric with worldwide validity.

3. Results & Discussion

The impact of geographic differences of memory colors on the evaluation of the color rendition properties of white light sources has been investigated in two ways. A first one examines the impact on the R_m,reg index values and a second one the impact on the predictive performance of the different regional memory color rendition metrics. For the former, respectively the latter, a database of 401 light sources—composed of commercially available sources, CIE illuminants, LED spectra, fluorescent spectra, etc.— [25] and a database of psychophysical data on subjective aspects—visual appreciation and naturalness—of color rendition were used.

3.1 Impact of geographic differences on R_m,reg index values

The impact on the R_m,reg index values was evaluated by comparing the regional values with those of the global index for 401 light source spectra. Agreement was assessed by means of the coefficient of determination R², the Spearman rank correlation coefficient r_S, the Mean-Absolute-Error and the STRESS measure. The results are given in Table 1 The agreement is also graphically illustrated in Fig. 3.

Table 1. Measures of agreement between the regional R_m,reg index scores (reg: BE, HU, BR, CO, TW, CN, IR) and the global R_m,reg index scores (reg: GL) for a set of 401 light sources [25].

View Table | View all tables in this article

Fig. 3 Comparison between the regional R_m,reg index scores (reg: BE, HU, BR, CO, TW, CN, IR) and the global R_m,reg index scores (reg: GL) for a set of 401 light sources.

Download Full Size | PDF

From the very high R² and r_S (resp. ≥ 93% and ≥96%, with the exception for BRazil) and very low STRESS values (≤ 2%) in Table 1, as well as the near 45° slopes observable in Fig. 3, it is clear that generally the regional metrics are highly correlated to the global metric. The near 45° slopes indicate that differences in visual color rendition are very similar across geographic regions, suggesting that the different metrics are likely to predict similar light source rank order, as was also confirmed by the high values of the Spearman rank correlation coefficient. However, as is also clear from Fig. 3, some index values do show, an albeit small, offset from the 45° line. This indicates, that although color rendition differences might be the same, their absolute level is not. These differences in absolute level were assessed by the MAE. From Table 1, it is clear that they are rather small, with the exception perhaps of that of TaiWan and IRan which show MAE values slightly larger than 5 units. It is commonly accepted that a 5 unit difference on a CIE-CRI-like scale corresponds to a just visible effect. Therefore, the results suggest that a single, i.e. the global, index can be used when relative light source rank order is the only parameter of interest, while it will result in small errors in predicted absolute color rendition level for some geographic regions. Although these small absolute differences could in principle be used to fine-tune the spectral design of light sources to specific geographic regions, the practical importance of these small errors can be questioned, as individual color appearance rating has been found to have a within-region variability of the same order or larger than the between-region variability [22].

3.2 Impact of geographic differences on R_m,reg predictive performance

The impact on the predictive performance of the R_m,reg metrics was investigated using a meta-correlation analysis identical to the one described by Smet et al. [20]. In that analysis, the Spearman rank correlation coefficients between the metric values of several light source sets and their observer ratings obtained in psychophysical experiments were calculated. Metric performance was assessed as the weighted average correlation coefficient corrected for sampling error and study artefacts according to the method of Hunter-Schmidt [26]. The following artefacts were corrected: study heterogeneity, range restriction/enhancement, inter-rater idiosyncrasy and sample correlation bias. The performance was investigated in terms of two subjective aspects of color rendition: visual appreciation (preference) and naturalness. Visual data were obtained for these two aspects from respectively twenty-one [27–41] and fifteen [27, 28, 30–34, 36, 37, 41–43] psychophysical experiments described in literature.

The results of the analysis are shown in Table 2 and Fig. 4. For comparison, the results for the CIE R_a [14] and Smet’s R_m (based on real objects) [12] metrics are shown as well.

Table 2. Metric performances (i.e. weighted average artefact corrected Spearman correlations, <r_S> ) for the subjective aspects of visual appreciation and naturalness.

View Table | View all tables in this article

Fig. 4 Performance of the regional MCRIs evaluated as the weighted average artefact corrected Spearman correlation coefficient in terms of visual appreciation and naturalness. The performance of the CIE R_a and Smet’s R_m are shown as well.

Download Full Size | PDF

For visual appreciation, the different regional metrics have a good predictive performance (<r_S> ≥ 80% for all metrics with the exception of HUngary and BRazil, median <r_S> = 85%). However, the performance is slightly lower than that of the Smet’s R_m. This could be due to differences in the experimental setup. The visual rating data on which the regional R_m,reg metrics are based were obtained on monitors, while the R_m is based on data obtained using real objects. It is known that visual perception on self-luminous displays is different from that using real objects [44, 45], causes being, among others, a reduced adaptation of the visual system to the former [45, 46] and the influence of surface texture, shading, motion parallax and binocular disparity on similarity ratings [47] and on colour constancy [48, 49].

For naturalness, the correlation is lower but still moderate (median <r_S> = 72%) and of approximately the same level as the regular MCRI. For a detailed discussion on reasons for the differences in performance for visual appreciation and naturalness, the reader is referred to [20].

From Fig. 4 it is clear that the regional R_m,reg metrics have a comparable performance for both subjective aspects of color rendition, with the exception of HUngary and BRazil, which have a slighly lower correlation.

To determine the statistical significance of these differences, the confidence-interval-method of Zou [50] for dependent overlapping correlations was applied under the null hypothesis of equal corrected average correlations, H₀: $\bar{r_{c}_{i}} - \bar{r_{c}_{j}} = 0$ (whereby i and j run over all metrics). A typical significance level of α = 0.05 was selected a priori. The method tests whether the 100 × (1- α)%-confidence interval (CI) contains zero, in which case H₀ cannot be rejected. The CI-bounds closest to zero, henceforth referred to as CI0 of the confidence interval tests, are summarized in Table 3. When there is a statistically significant difference, the CI0 has been underlined and marked in red.

Table 3. Results for the cross-comparisons of metric performance (H₀: $\bar{r_{c}_{i}} - \bar{r_{c}_{j}} = 0$ ). The values are the confidence interval bound closest to zero. Red underlined values signify statistically significant differences, i.e. zero was within the CI bounds. Results for visual appreciation and naturalness are respectively shown in the upper and lower triangles of the table.

View Table | View all tables in this article

From Table 3 it can be seen that all memory color rendition metrics had significantly higher correlation coefficients than the CIE R_a index for visual appreciation. For naturalness, no significant differences were observed with the CIE R_a index.

For visual appreciation, the global R_m,reg index was only significantly different from the index for HUngary and BRazil. The latter two were also significantly different from one another, although only barely as can be seen from the small magnitude of CI0. Other, although small significant differences, were between (CN,BE), (CN,BR), (CN,IR) and (TW,IR).

For naturalness, the following small, but significant, differences were found: (IR, GL), (IR, TW), (CN, CO), (TW,BR), (TW,BE), (IR,CN), (TW,HU) and (CO, GL).

Overall, and in agreement with the earlier conclusion about metric value differences being approximately independent of geographic region, the results show that for most cross-comparisons the impact of geographic region on the predictive performance is not or only barely significant.

Note that, again, the same remark can be made about the significance of the results of the region average observer for those to be expected for an individual observer.

4. Conclusions

In a recent study [22], the effect of cross-regional variation on color appearance rating and memory colors of eleven familiar objects was investigated in seven different geographic regions: Belgium, Hungary, Brazil, Colombia, Taiwan, China and Iran.

Based on the bivariate Gaussian models of the observer rating data obtained by Smet et al. [22], seven regional and one global memory color rendition metric have been created. Using the latter, the impact of cross-regional differences in color appearance rating on color rendition evaluation was examined.

In a first analysis, the regional R_m,reg index values for a database of 401 light sources have been compared with those obtained using the global metric. The results indicate that the regional metrics are highly correlated with the global one. In addition, plots of the regional index values versus the global index values all have slopes close to 45°, indicating that color rendition differences between light sources are approximately preserved across regional metrics. However, the plots do show an offset indicating that absolute level is not preserved, although the Mean-Absolute-Errors between the regional and the global index values were small for most geographic regions. With the exception of TaiWan and IRan, all differences were smaller than 5 scale units. It is commonly accepted that a 5 unit difference on a “CIE CRI”-like scale corresponds to a just visible effect. Therefore, the results suggest that a single global index can be used when relative light source rank order is the only parameter of interest, while it will likely result in small errors in predicted absolute color rendition level for some geographic regions.

A second analysis examined the impact of cross-regional differences on the predictive performance of the regional memory color rendition metrics. Performance was assessed as the artefact corrected weighted average Spearman correlation—calculated according the method of Hunter-Schmidt—between metric predictions and psychophysical data published in literature. Predictive performance with respect to two subjective aspects of color rendition was investigated: visual appreciation (preference) and naturalness. Data of respectively twenty-one and fifteen experiments from seventeen studies were used. The results showed that most of the regional memory color rendition metrics had a high and comparable performance (median <r_S> = 85%) for visual appreciation. Compared to the Smet’s MCRI R_m, which is based on color appearance rating data obtained using real objects, the performance is lower. However, this is likely due to differences in perception between real object presentation and presentation on self-luminous displays, for example due to reduced adaptation to the latter or due to the differences in surface texture, shading, motion parallax and binocular disparity, all of which are known to have an effect on similarity ratings and color constancy. For naturalness, the regional memory colour rendition metrics had a moderate, but comparable performance (median <r_S> = 72%). A cross-comparison of the metric performance using the confidence interval method of Zou showed that for most of them the impact of geographic region on the predictive performance is not or only barely significant (p < 0.05).

Note that the conclusions listed above are applicable when dealing with color rendition, as expressed by the concept of an average observer. As was shown by Smet et al. [22], individual memory color based color appearance ratings have a within-region variability of the same order or larger than the between-region variability. Therefore, within-region differences—between individual observers and that of the region average observer—in color rendition evaluation are expected to be of similar magnitude as the ones reported above.

To conclude, the analyses indicates that a single, globally valid memory color rendition metric can be proposed which predicts light source rank order well and which results in only minor errors—of the same magnitude as those induced by inter-observer variability—in the absolute level of the predicted color rendition for some geographic regions.

In addition, as the global and Belgian R_m,reg metrics are very comparable, Smet’s R_m [12], obtained using Belgian test subjects (but different from the ones that participated in the experiments described in [22]), would be a good approximation to that globally valid metric, while ensuring a higher predictive performance due to the more representative (for color rendition evaluation) object presentation adopted during its development.

Such a globally valid metric—use Smet’s R_m for best performance—cannot only be applied to evaluate the color rendition properties of existing light sources but can also be implemented in the spectral optimization of new light sources that have a good overall (globally) lighting quality (visual appreciation). And although region-specific fine-tuning of light source spectra could in principle be done using the similarity function published in [22], the size of the cross-regional differences, comparable to the typical within-region observer variability, suggest that such spectral optimization might be of little practical value. Furthermore, lighting designers should bare in mind that those similarity functions were derived for objects presented on a monitor, resulting in a lower predictive performance than Smet’s R_m.

Acknowledgments

Author KS would like to thank the Research Foundation Flanders (FWO) for supporting this study through a Postdoctoral Fellowship (12B4913N).

References and links

1. Hering, Grundzüge der Lehre vom Lichtsinn (Springer-Verlag, 1920).

2. C. J. Bartleson, “Color in memory in relation to photographic reproduction,” Photo. Sci. Eng. 5, 327–331 (1961).

3. S. N. Yendrikhovskij, F. J. J. Blommaert, and H. de Ridder, “Color reproduction and the naturalness constraint,” Color Res. Appl. 24(1), 52–67 (1999). [CrossRef]

4. P. Bodrogi and T. Tarczali, “Colour memory for various sky, skin, and plant colours: Effect of the image context,” Color Res. Appl. 26(4), 278–289 (2001). [CrossRef]

5. P. Bodrogi and T. Tarczali, “Investigation of Colour Memory,” in Colour Image Science: Exploiting Digital Media, L. W. MacDonald, and M. R. Luo, eds. (John Wiley & Sons Limited, 2002), pp. 23–48.

6. H. Zeng and R. Luo, “Modelling memory colour region for preference colour reproduction,” in SPIE: Color Imaging XV: Displaying, Processing, Hardcopy, and Applications (2010), pp. 752808.

7. T. Yano and K. Hashimoto, “Preference index for japanese complexion color under illumination,” J. Light Vis. Env. 22, 269–274 (1998).

8. C. Boust, H. Brettel, F. Viénot, S. Berche, and G. Alquié, “Color enhancement of digital images by experts and preference judgments by observers,” J. Imaging Sci. 50(1), 1–11 (2006). [CrossRef]

9. S. Xue, M. Tan, A. McNamara, J. Dorsey, and H. Rushmeier, “Exploring the use of memory colors for image enhancement,” in SPIE: Human Vision and Electronic Imaging XIX(2014), pp. 901411.

10. D. B. Judd, “A flattery index for artificial illuminants,” J. Illum. Eng. 62, 593–598 (1967).

11. C. L. Sanders, “Assessment of color rendition under an iIlluminant using color tolerances for natural objects,” J. Illum. Eng. 54, 640–646 (1959).

12. K. A. G. Smet, W. R. Ryckaert, M. R. Pointer, G. Deconinck, and P. Hanselaer, “A memory colour quality metric for white light sources,” Energy Build. 49, 216–225 (2012). [CrossRef]

13. W. A. Thornton, “A validation of the color-preference index,” J. Illum. Eng. 4(1), 48–52 (1974). [CrossRef]

14. CIE13, 3–1995, “Method of measuring and specifying colour rendering properties of light sources,” (CIE, Vienna, Austria, 1995).

15. P. Bodrogi, S. Brückner, and T. Q. Khanh, “Ordinal scale based description of colour rendering,” Col. Res. Appl. 36(4), 272-285 (2010).

16. W. Davis and Y. Ohno, “Color quality scale,” Opt. Eng. 49(3), 033602 (2010). [CrossRef]

17. C. Li, M. R. Luo, and C. Li, “Assessing colour rendering properties of daylight sources part ii: a new colour rendering index: CRI-CAM02UCS,” (2009).

18. K. A. G. Smet, J. Schanda, L. Whitehead, and R. M. Luo, “CRI2012: A proposal for updating the CIE colour rendering index,” Lighting Res. Tech. 45(6), 689–709 (2013). [CrossRef]

19. A. Zukauskas, R. Vaicekauskas, F. Ivanauskas, H. Vaitkevicius, P. Vitta, and M. S. Shur, “Statistical approach to color quality of solid-state lamps,” IEEE J. Sel. Top. Quantum Electron. 15(6), 1753–1762 (2009). [CrossRef]

20. K. Smet and P. Hanselaer, “Memory and preferred colours and the colour rendition of white light sources,” Lighting Res. Technol., in press (2015).

21. K. A. G. Smet, W. R. Ryckaert, M. R. Pointer, G. Deconinck, and P. Hanselaer, “Colour appearance rating of familiar real objects,” Color Res. Appl. 36(3), 192–200 (2011). [CrossRef]

22. K. A. G. Smet, Y. Lin, B. V. Nagy, Z. Németh, G. L. Duque-Chica, J. M. Quintero, H.-S. Chen, R. M. Luo, M. Safi, and P. Hanselaer, “Cross-cultural variation of memory colors of familiar objects,” Opt. Express 22(26), 32308–32328 (2014). [CrossRef] [PubMed]

23. E. Marszalec, B. Martinkauppi, M. Soriano, and M. Pietikäinen, “A physics-based face database for color research,” J. Electron. Imaging 9(1), 32–38 (2000). [CrossRef]

24. Q. Sun and M. D. Fairchild, “Statistical characterization of face spectral reflectances and its application to human portraiture spectral estimation,” J. Imaging Sci. Technol. 46, 498–506 (2002).

25. K. W. Houser, M. Wei, A. David, M. R. Krames, and X. S. Shen, “Review of measures for light-source color rendition and considerations for a two-measure system for characterizing color rendition,” Opt. Express 21(8), 10393–10411 (2013). [CrossRef] [PubMed]

26. J. E. Hunter and F. L. Schmidt, Methods of Meta-Analysis:Correcting Error and Bias in Research Findings (Sage Publications, Inc, 2004).

27. S. Jost-Boissard, M. Fontoynont, and J. Blanc-Gonnet, “Thurstone scalings for attractiveness of fruit and vegetables under nine 3000K and eight 4000K light sources. (personal communication),” (2009).

28. S. Jost-Boissard, M. Fontoynont, and J. Blanc-Gonnet, “Perceived lighting quality of LED sources for the presentation of fruit and vegetables,” J. Mod. Opt. 56(13), 1420–1432 (2009). [CrossRef]

29. N. Narendran and L. Deng, “Color rendering properties of LED light sources,” Proc. SPIE 4776, 61–67 (2002). [CrossRef]

30. J. Schanda and G. Madár, “Light Source Quality Assessment,” in CIE 26th Session 2007(CIE, 2007), pp. 72–75.

31. F. Szabó, P. Csuti, and J. Schanda, “Colour preference under different illuminants – new approach of light source colour quality,” in Light and Lighting Conference with Special Emphasis on LEDs and Solid State Lighting (CIE, 2009), pp. PWDAS-43.

32. J. Vanrie, “Appendix 4: Technical report to the user committee of the IWT-TETRA project (80163): the effect of the spectral composition of a light source on the visual appreciation of a composite objectset,” (PHL, 2009), p. 11.

33. K. A. G. Smet, W. R. Ryckaert, M. R. Pointer, G. Deconinck, and P. Hanselaer, “Memory colours and colour quality evaluation of conventional and solid-state lamps,” Opt. Express 18(25), 26229–26244 (2010). [CrossRef] [PubMed]

34. A. Tsukitani and Y. Lin, “Research on FCI (presented at the meeting of TC1-91 in Kuala Lumpur, Malaysia, April 23-26, 2014),” (2014).

35. Y. Imai, T. Kotani, and T. Fuchida, “A study of color rendering properties based on color preference in adaptation to LED lighting ” in CIE2012 Lighting Quality & Energy Efficiency (CIE, 2012), pp. 369–374.

36. K. W. Houser, D. K. Tiller, and X. Hu, “Tuning the fluorescent spectrum for the trichromatic visual response: a pilot study,” Leukos 1(1), 7–23 (2005). [CrossRef]

37. K. A. G. Smet, W. R. Ryckaert, M. R. Pointer, G. Deconinck, and P. Hanselaer, “Optimization of colour quality of LED lighting with reference to memory colours,” Lighting Res. Tech. 44(1), 7–15 (2012). [CrossRef]

38. Y. Ohno and W. Davis, “Visual evaluation experiment on chroma enhancement effects in color rendering of light sources, report submitted to TC1-69,” National Institute of Standards and Technology (2010).

39. M. Wei, K. W. Houser, G. R. Allen, and W. W. Beers, “Color preference under LEDs with diminished yellow emission,” Leukos 10(3), 119–131 (2014). [CrossRef]

40. Y. Imai, T. Kotani, and T. Fuchida, “A study of color rendering properties based on color preference of objects in adaptation to LED lighting,” in CIE Centenary Conference “Towards a New Century of Light” (CIE, 2013), pp. 62–67.

41. A. Tsukitani, “Optimization of colour quality for landscape lighting based on feeling of contrast index,” in CIE Centenary Conference “Towards a New Century of Light” (CIE, 2013), pp. 68–71.

42. M. S. Rea and J. P. Freyssinier, “Color rendering: beyond pride and prejudice,” Color Res. Appl. 35(6), 401–409 (2010). [CrossRef]

43. M. S. Rea and J. P. Freyssinier-Nova, “Color rendering: a tale of two metrics,” Color Res. Appl. 33(3), 192–202 (2008). [CrossRef]

44. R. W. G. Hunt, The Reproduction of Colour (John Wiley and Sons, 2004).

45. B. Oicherman, M. R. Luo, B. Rigg, and A. R. Robertson, “Adaptation and colour matching of display and surface colours,” Color Res. Appl. 34(3), 182–193 (2009). [CrossRef]

46. M. D. Fairchild, “Chromatic adaptation to image displays,” Proc. TAGA 2, 803–824 (1992).

47. S. N. Yendrikhovskij, F. J. J. Blommaert, and H. de Ridder, “Representation of memory prototype for an object color,” Color Res. Appl. 24(6), 393–410 (1999). [CrossRef]

48. R. S. Berns and M. E. Gorzynski, “Simulating surface colors on CRT displays: the importance of cognitive clues,” in AIC Colour and Light (1991), pp. 21–24.

49. D. H. Brainard, “Color constancy in the nearly natural image. 2. Achromatic loci,” J. Opt. Soc. Am. A 15(2), 307–325 (1998). [CrossRef] [PubMed]

50. G. Y. Zou, “Toward using confidence intervals to compare correlations,” Psychol. Methods 12(4), 399–413 (2007). [CrossRef] [PubMed]

Region:	BE	HU	BR	CO	TW	CN	IR
R²	0.98	0.93	0.76	0.93	0.97	0.97	0.99
*r_S*	0.99	0.97	0.87	0.96	0.98	0.98	0.99
*MAE*	1.82	2.35	3.11	4.63	8.39	1.77	5.75
*STRESS*	0.01	0.01	0.02	0.01	0.01	0.01	0.00

*<r_S>*	CIE R_a	*R_m*	R_m GL	*R_m* BE	R_m HU	R_m BR	R_m CO	R_m TW	R_m CN	*R_m* IR
Visual appreciation	0.09	1.00	0.86	0.86	0.76	0.61	0.89	0.84	0.93	0.81
Naturalness	0.60	0.71	0.71	0.65	0.75	0.76	0.58	0.84	0.73	0.58

*CI0*	CIE R_a	*R_m*	R_m GL	*R_m* BE	*R_m* HU	*R_m* BR	*R_m* CO	*R_m* TW	*R_m* CN	*R_m* IR
CIE R_a		-0.68	-0.45	-0.42	-0.34	-0.19	-0.43	-0.38	-0.51	-0.33
*R_m*	−0.28		−0.03	−0.03	0.03	0.13	−0.03	−0.05	−0.06	0.00
R_m GL	−0.28	0.28		−0.12	0.03	0.06	0.14	−0.10	0.09	−0.06
R_m BE	−0.35	0.19	0.00		−0.06	0.00	0.06	−0.04	-0.01	0.00
R_m HU	−0.22	−0.22	−0.16	−0.10		0.02	0.09	0.06	0.03	0.11
R_m BR	−0.19	−0.29	−0.25	−0.21	−0.15		0.02	0.00	-0.05	0.06
R_m CO	0.43	0.11	-0.01	0.02	0.09	0.20		−0.09	0.02	−0.02
R_m TW	−0.13	−0.11	−0.02	0.04	0.02	−0.13	0.05		0.03	0.00
R_m CN	−0.28	−0.21	−0.09	−0.01	0.23	0.31	0.08	0.09		0.02
R_m IR	0.41	0.16	-0.13	0.01	0.04	0.13	−0.10	-0.11	-0.03

Region:	BE	HU	BR	CO	TW	CN	IR
R²	0.98	0.93	0.76	0.93	0.97	0.97	0.99
*r_S*	0.99	0.97	0.87	0.96	0.98	0.98	0.99
*MAE*	1.82	2.35	3.11	4.63	8.39	1.77	5.75
*STRESS*	0.01	0.01	0.02	0.01	0.01	0.01	0.00

*<r_S>*	CIE R_a	*R_m*	R_m GL	*R_m* BE	R_m HU	R_m BR	R_m CO	R_m TW	R_m CN	*R_m* IR
Visual appreciation	0.09	1.00	0.86	0.86	0.76	0.61	0.89	0.84	0.93	0.81
Naturalness	0.60	0.71	0.71	0.65	0.75	0.76	0.58	0.84	0.73	0.58

Impact of cross-regional differences on color rendition evaluation of white light sources

Abstract

1. Introduction

2. Regional Memory Color Rendition Indices, R_m,reg

2.1 Similarity functions

2.2. Spectral reflectance

2.3 Calculation of the general similarity indicator, S_a,reg

2.4 Rescaling the S_a,reg indicator to the R_m,reg index

3. Results & Discussion

3.1 Impact of geographic differences on R_m,reg index values

3.2 Impact of geographic differences on R_m,reg predictive performance

4. Conclusions

Acknowledgments

References and links

Cited By

Figures (4)

Tables (3)

Equations (2)

Optics Express

Abstract

1. Introduction

2. Regional Memory Color Rendition Indices, Rm,reg

2.1 Similarity functions

2.2. Spectral reflectance

2.3 Calculation of the general similarity indicator, Sa,reg

2.4 Rescaling the Sa,reg indicator to the Rm,reg index

3. Results & Discussion

3.1 Impact of geographic differences on Rm,reg index values

3.2 Impact of geographic differences on Rm,reg predictive performance

4. Conclusions

Acknowledgments

References and links

Cited By

Figures (4)

Tables (3)

Equations (2)

Optics Express

2. Regional Memory Color Rendition Indices, R_m,reg

2.3 Calculation of the general similarity indicator, S_a,reg

2.4 Rescaling the S_a,reg indicator to the R_m,reg index

3.1 Impact of geographic differences on R_m,reg index values

3.2 Impact of geographic differences on R_m,reg predictive performance