The appreciation of LED-based white light sources by Dutch and Chinese people in three application areas

Abstract

The aim of this study was to evaluate differences in appreciation for LED-based white-light sources between Dutch and Chinese people, when used for the illumination of three applications: fresh food, packaging material and skin tone. Furthermore, the contribution of the CIE special and general colour rendering indices (R_i and R_a), a colour gamut measure (G_a), and chroma changes for specific test-colour samples to perceived attractiveness was investigated. Thirty-four Dutch and 36 Chinese people assessed the attractiveness of the object appearance, with paired comparison experiments, for seven light sources at two CCT settings, 3000 K (Dutch and Chinese) and 4000 K (Chinese only), and a range of R_a and G_a values. It was found that for illuminating fresh food and packaging material, most Dutch and Chinese participants preferred light settings with an increased colour gamut. In contrast to colour rendering indices, the chroma change for the strong red test colour sample could be used to predict object attractiveness, but the established equations were different per application and culture. There was no clear relation between objective measures and skin tone preference for the Dutch study. Finally, Chinese participants did not like an increase in red saturation at 3000 K, but they allowed an increase at 4000 K.

1. Background

1.1. Limitations of colour fidelity indices

Colour rendering is defined as the effect of an illuminant on the colour appearance of objects by conscious or subconscious comparison with their colour appearance under a reference illuminant.¹ The general colour rendering index (CRI-R_a)² is used to measure and specify the colour rendering ability of a white light source, based on a set of eight specific CIE 1974, moderately saturated, test-colour samples (TCS). The procedure for calculating R_a is described in CIE13.3-1995.³ Although R_a has been used for almost 50 years, it is well known that there are some limitations (see e.g. Houser et al.⁴). Recently, CIE TC1-90 published their research report with the new CIE 2017 fidelity index for accurate scientific use, CIE-R_f.⁵ This index is based on the fidelity index of IES TM-30-15^6–8 (IES-R_f), but has a different scaling factor, slightly modified TCS, and another CCT transition range for the reference illuminant from the Plankian radiator to a phase of daylight. Like R_a, CIE-R_f is an index that combines all computed colour differences into one average value and is only one aspect of colour quality. Earlier studies have already concluded that newer indices, purely based on colour fidelity, are not expected to provide more information compared to the current one.^9,10 Therefore, the new fidelity index is to be used for scientific purposes only and is not recommended by the CIE as a replacement for R_a, neither for the purpose of rating and specification of products nor for regulatory or other minimum performance requirements. Consequently, the current CIE general colour rendering index, R_a, will continue to be used as the global standard for measuring and specifying colour fidelity.

Although sometimes wrongly assumed, a light source with a high R_a value is not always most preferred.^11–17 Fidelity is only one aspect of colour rendition and does not disclose the direction of the colour shifts. Fidelity can be reduced equally due to an increase or decrease in chroma. To address this limitation, Guo and Houser,¹⁸ Houser et al.,¹⁰ Rea and Freyssinier-Nova¹⁹ and Freyssinier and Rea²⁰ advocate the use of a two-dimensional system based on colour fidelity and colour gamut to more completely, but still objectively, describe the colour rendition properties of white-light sources. Such a two-dimensional system also shows the trade-off between maximising colour fidelity and enlarging the gamut area, because it is impossible to simultaneously reach a high colour fidelity and a high colour saturation, as increasing saturation would sacrifice the fidelity.²¹ Indeed, Jost-Boissard et al.^12,13 found that naturalness is better described by fidelity indices and colourfulness and attractiveness are better described by gamut-based indices. Teunissen et al.¹⁶ also found that light sources resulting in increased gamut areas were more preferred compared to light sources with high fidelity or a reduced gamut area. Similarly, Ohno et al.¹¹ found that preference is mainly affected by the perceived colourfulness of objects, with a higher preference for a more colourful appearance. However, it should be noted that the increase in colourfulness should not be too high otherwise the preference starts decreasing.¹¹ The change in colourfulness can be expressed by an average change in chroma. This measure is used by Khanh et al.^22–24 who found a relation between the mean chroma shift and preference.

Although the combination of colour fidelity and colour gamut, or an average change in chroma, better describes differences in preference, the shape of the colour gamut for the test source, relative to the colour gamut for reference illuminant, also contributes to preference.²⁵ This indicates that chroma changes for some specific colours may be more important than for others. Information on colour shifts for the individual TCS might therefore be necessary to explain differences in preference.

To overcome the limitation of fidelity indices for ranking light-sources in order of preference, gamut based indices and chroma-changes for individual TCS can be used. The next two sections describe the objective measures used in this study.

1.2. A simple CRI-based colour gamut index

Teunissen et al.¹⁶ and Teunissen and Hoelen¹⁷ proposed using a relative gamut area index (G_a), which is also referred to as the colour saturation index (CSI), in addition to the CIE general colour rendering index (R_a) to indicate the change in gamut area for the white-light source in comparison to the reference illuminant. The process for calculating G_a is described by Teunissen and Hoelen¹⁷ and based on the rating procedure described in Section 5 of CIE publication 13.3-1995.³ The gamut area for the reference illuminant (G_r) and the gamut area for the lamp to be tested (G_k) shall be calculated with the chromaticity coordinates for the first eight CIE 1974 TCS, derived respectively according to Sections 5.6 and 5.7 in CIE13.3-1995.³ The gamut areas for the reference illuminant (G_r) and for the test source (G_k) are computed according to equations (1) and (2), respectively.

\begin{matrix} G_{r} = ((u_{r, 1} \times v_{r, 2}) - (u_{r, 2} \times v_{r, 1}) + (u_{r, 2} \times v_{r, 3}) \\ - (u_{r, 3} \times v_{r, 2}) + (u_{r, 3} \times v_{r, 4}) - (u_{r, 4} \times v_{r, 3}) \\ + (u_{r, 4} \times v_{r, 5}) - (u_{r, 5} \times v_{r, 4}) + (u_{r, 5} \times v_{r, 6}) \\ - (u_{r, 6} \times v_{r, 5}) + (u_{r, 6} \times v_{r, 7}) - (u_{r, 7} \times v_{r, 6}) \\ + (u_{r, 7} \times v_{r, 8}) - (u_{r, 8} \times v_{r, 7}) + (u_{r, 8} \times v_{r, 1}) \\ - (u_{r, 1} \times v_{r, 8})) / 2 \end{matrix}

(1)

\begin{matrix} G_{k} = ((u'_{k, 1} \times v'_{k, 2}) - (u'_{k, 2} \times v'_{k, 1}) \\ + (u'_{k, 2} \times v'_{k, 3}) - (u'_{k, 3} \times v'_{k, 2}) \\ + (u'_{k, 3} \times v'_{k, 4}) - (u'_{k, 4} \times v'_{k, 3}) \\ + (u'_{k, 4} \times v'_{k, 5}) - (u'_{k, 5} \times v'_{k, 4}) \\ + (u'_{k, 5} \times v'_{k, 6}) - (u'_{k, 6} \times v'_{k, 5}) \\ + (u'_{k, 6} \times v'_{k, 7}) - (u'_{k, 7} \times v'_{k, 6}) \\ + (u'_{k, 7} \times v'_{k, 8}) - (u'_{k, 8} \times v'_{k, 7}) \\ + (u'_{k, 8} \times v'_{k, 1}) - (u'_{k, 1} \times v'_{k, 8})) / 2 \end{matrix}

(2)

Finally, the CSI (G_a) is derived from G_k and G_r by using the following formula:

G_{a} = 100 \times \frac{G_{k}}{G_{r}}

(3)

1.3. CIELUV-based relative chroma changes

Unfortunately, chroma shifts for the individual TCS cannot be computed with the colour spaces used for CRI computations, i.e., CIE 1960 u,v and CIE 1964 W*U*V*. The CIE 1976 L*u*v* (CIELUV) colour space is closest related to the CRI used colour spaces and therefore is used to compute the chroma changes of the test source relative to those of the reference illuminant for the 14 CIE 1974 defined TCS,³ as well as for an additional TCS (#15) for the East-Asian woman’s complexion, as defined in JIS Z 8726-1990.²⁶ The spectral radiance factors for the saturated TCS (#09-12) as well as for the skin tones (TCS 13 and 15) are included in Figure 1.

First, the chromaticity coordinates for the reference illuminant (r) and the chromatically adapted chromaticity coordinates for the test source (k) need to be transformed to the CIE 1976 u′,v′ coordinates. Subsequently, for all TCS (i) and for both the test source and the reference illuminant, the CIELUV chroma ( $C_{k / r, i}^{*}$ ) can be computed, see clause 4.3 in ISO/CIE 11664-5 ²⁷:

\begin{matrix} CIE 1976 u, v (CIELUV) chroma : \\ C_{k / r, i}^{*} = 13 L_{k / r, i}^{*} \sqrt{(u'_{k / r, i} - u'_{n})^{2} + (v'_{k / r, i} - v'_{n})^{2}} \end{matrix}

(4)

The lightness component ( $L_{k / r, i}^{*}$ ) can be computed according to equation (10) in ISO/CIE 11664-5²⁷ and the white object colour stimulus values $(Y_{n}, u'_{n}, v'_{n})$ are derived from the reference illuminant, assuming a perfect white diffusor (Y_n = 100).

Finally, the relative chroma change for all TCS (i) can be computed according to the following equation:

\begin{matrix} Relativechromachange (%) : \\ C_{i} = 100 \cdot (\frac{C_{k, i}^{*}}{C_{r, i}^{*}} - 1) \end{matrix}

(5)

As described in Section 1.1, the relative chroma changes for some specific TCS may contribute to the appreciation of the object appearance.

1.4. Aim of the current study

Teunissen et al.¹⁶ and Teunissen and Hoelen¹⁷ conducted a study in the Netherlands to investigate the impact of colour fidelity (R_a) and colour gamut (G_a) on the user preference for white-light sources in three application areas. It was found that the halogen light source, with highest colour fidelity (R_a = 100), was not preferred for any of the three application areas. LED-based white-light sources with a larger gamut area (G_a > 105) were most preferred. The results revealed that other than colour fidelity, colour saturation makes a significant contribution to the appreciation of a white-light source. However, the results were application dependent and gender specific.

The above study was conducted in the Netherlands with Dutch people, and to further explore a possible cultural difference, it was decided to repeat this study in Shanghai with Chinese participants. In the Dutch study, all experiments were conducted with light settings at 3000 K, whereas in China, the correlated colour temperature is typically 4000 K. To check cultural differences and the influence of CCT on user preference, the experiments in China were conducted with light settings at both 3000 K and 4000 K. This paper mainly describes the experiments conducted in China, but also compares the results to the Dutch study. In addition, chroma changes for the individual TCS are used to explain the appreciation for a light setting.

2. Colour rendering preference study for LED-based white-light sources in China

2.1. Set-up

To make the comparison between cultures, we reproduced the experimental environment and the seven tested light settings used in the Dutch study, in the research facilities in Shanghai. Two light boxes with dimensions of 60 cm (wide) × 60 cm (deep) × 60 cm (high) and Telelumen light replicators mounted at the top were used to generate the light settings. The two boxes were placed side by side and the inside of the boxes was covered with matte black textile to minimise the surface reflections. The Telelumen light replicator is a LED luminaire with 16 independently controllable colour channels, which enables users to create a wide range of emission spectra. The seven light settings used in the Dutch study were reproduced with the Telelumen light replicators to obtain the same R_a–G_a combinations. The spectral power distributions (SPDs) used in the Chinese study are illustrated in Figure 2. The SPD name consists of the information for CCT, R_a and G_a values. For setting ‘3R80G100’, for example, the ‘3’ indicates a CCT of 3000 K and ‘R80G100’ represents an approximate R_a value of 80 and G_a value of 100. Table 1 lists the detailed parameters for the test light settings, including the colour coordinates (u′,v′), correlated colour temperature, the general colour rendering index (R_a), the CSI (G_a) and the special colour rendering indices (R_i) and the relative chroma changes (C_i) for specific TCS. TCS9 to TCS12 represent strong red, yellow, green, and blue, respectively. TCS13 is for Caucasian skin tones and TCS15 is for East-Asian woman’s complexion.²⁷ The spectral radiance factors for these TCS are included in Figure 1. The values for TCS 14 (green leaf) are not included because this sample was not considered as relevant for this study. The R_a values of the seven light settings vary from 72 to 99, while the G_a values vary from 89 to 122. In the Dutch study, one of the tested light sources was a halogen light source with both R_a and G_a values equal to 100. But due to the limitation of the Telelumen light replicator, it was not possible to exactly replicate this setting. The closest match resulted in a light setting with an R_a value of 98 and a G_a value of 102. The correlated colour temperature of the seven light settings was on average 3007 K (SD = 23 K), and the mean illuminance, measured at the object surface, was 738 lux (SD = 18 lux).

Table 1

Colorimetric properties of the 3000 K light settings.

SPD	3R100G100	3R93G100	3R80G100	3R80G110	3R80G115	3R80G90	3R70G120
CIE1976-u'	0.2510	0.2490	0.2497	0.2489	0.2513	0.2497	0.2499
CIE1976-v'	0.5231	0.5232	0.5228	0.5229	0.5228	0.5217	0.5222
CCT (K)	2978	3028	3012	3031	2973	3019	3012
D _uv	0.001	0.002	0.001	0.002	0.001	0.001	0.001
R _a	98	93	81	83	81	79	72
G _a	102	101	98	112	114	91	118
R₉	92.9	66.6	14.2	39.8	44.1	−23.1	8
R ₁₀	94.8	87.7	63.3	96.6	90.4	79.7	78.2
R ₁₁	96.6	92.9	79.6	71.4	58	87.5	43.5
R ₁₂	90.4	86.6	55.5	94.6	90.2	77.4	77.3
R ₁₃	99.8	94.7	82.9	84.2	81.2	81.7	70.6
R ₁₅	98.5	86.6	75.3	78.1	81.4	58.8	67.6
C ₉	−1.2	−5.6	−14.4	10	9.3	−20.7	15.3
C ₁₀	−1.2	−2.5	−5.9	0.5	2.5	−3.6	5.4
C ₁₁	1.2	2.7	−7	12.2	18	−4.4	24.1
C ₁₂	−0.7	−3.1	−4.6	2.6	0.5	−2.2	2.1
C ₁₃	0.1	−2.6	−8.5	8.1	10.1	−8.9	15.7
C ₁₅	−0.7	−6.5	−12.1	11.1	9.5	−20.4	16.5

Another seven SPDs with CCTs of 4000 K were created to obtain similar R_a–G_a combinations. The illuminance of these seven light settings was on average 708 lux (SD = 8 lux). Figure 3 and Table 2 show the SPDs and R_a, G_a, R_i values and chroma changes (C_i) for the seven light settings at 4000 K.

Table 2

Colorimetric properties of the 4000K light settings.

SPD	4R100G100	4R93G100	4R80G100	4R80G110	4R80G115	4R80G90	4R70G120
CIE1976-u'	0.2253	0.2252	0.2251	0.2252	0.2253	0.2250	0.2252
CIE1976-v'	0.5017	0.5018	0.5018	0.5015	0.5016	0.5018	0.5017
CCT (K)	3990	3995	3997	3999	3994	3999	3996
D _uv	−0.0001	0.0001	0.0001	−0.0001	0.0000	0.0002	0.0000
R _a	99	92	80	83	82	80	72
G _a	100	101	96	112	114	89	122
R₉	86.5	71.7	17.1	23.6	27.5	−43.9	−13.9
R ₁₀	99.1	81.9	56.8	94.2	90	81.8	91.3
R ₁₁	97.7	92.9	77.3	75.5	63.2	91.7	51.4
R ₁₂	98.8	78.8	51.2	91.2	94.4	82.3	90.7
R ₁₃	99.5	94.1	81.9	84.5	82.2	83.1	75.4
R ₁₅	97.5	90.2	77.7	75.1	78.1	56.9	64
C ₉	−2.4	−5	−14.7	13.5	12.8	−25.6	20.1
C ₁₀	0	−0.7	−1	0.4	0.9	−0.6	1
C ₁₁	1	1.6	−6.9	11	16.6	−1.9	22
C ₁₂	0	0.4	1.4	4.1	0.7	−1.3	−2.9
C ₁₃	0.3	−2.4	−7.3	6.4	7.8	−6	10.8
C ₁₅	−1.2	−4.8	−11.1	13	11.8	−21.5	19.7

During the experiments, the SPDs with the same CCT were compared, but no cross-CCT effects were investigated in this study.

2.2. Method

A paired-comparison method was used in this study, with a three-point numerical scale to indicate the difference in attractiveness of the displayed objects, being slightly more (1), more (2), and much more (3) attractive. The experiment was divided into two separate experiments, one for 3000 K and the other for 4000 K. In each experiment, there were three sessions addressing the three applications: fresh food, packaging material and skin tone, as shown in Figure 4. Fresh food consisted of real fruits (lemons, apples, bananas, peaches) and vegetables (an aubergine), and packaged food included some packaged biscuits, cans of chips, a can of cola, a box of teabags, a box of coffee powder and a box of blueberry milk. Both applications covered a range of colours, like red, green, yellow, blue and so on. For the skin tone application, the respondents’ own hands were used.

Identical items were placed on the bottom surface of the boxes (in the skin tone application, the respondent’s left hand was put in the left box and the right hand was put in the right box), but each box was illuminated with different SPDs. It must be noted that the objects used in the Chinese study are familiar to Chinese people and therefore slightly different compared to the familiar objects used in the Dutch study.¹⁶

The experiments, one for 3000 K and the other one for 4000 K, were conducted in an otherwise dark room. For each application, the participants were seated in front of the boxes and instructed to indicate on which side the objects looked more attractive using the 3-point scale. In addition, the participants were also asked to describe the reason for their selections orally. The presentation order of the comparisons was randomised over the applications and over the participants. The order of the applications fresh food and packaging material was also randomised over the participants. To minimise the impact of external factors, like outdoor temperature preceding the experiment, skin tone was always evaluated in the last session. For each application and both CCTs, the participants evaluated 25 SPD pairs, including 4 introductory pairs and 21 formal pairs. The four introductory SPD pairs were included for each application to help the participants become familiar with their task and the possible variations in object appearance, but the introductory SPD pairs were not used in the data analysis. In total, each participant made (3 applications × 2 CCTs × 25 pairs=) 150 paired comparisons.

2.3. Participants

Thirty-six Chinese volunteers, 19 males and 17 females, participated in this study. They were recruited by an external agency and were not employed by, or working for, Philips. Their ages ranged from 26 to 53 years with an average of age 40 years (SD = 8.4 years). Before starting the experiment, each of the participants had to pass the vision test (GB11533-2011²⁸) and colour blindness test, for the latter using the Yu Ziping colour blind chart.²⁹

3. Results for the Chinese study

The data of all 36 participants were collected for the 3000 K experiment, but two persons did not participate in the second (4000 K) experiment. So, the data for the remaining 34 participants were used for the analysis of the light settings at 4000 K.

There was no direct comparison between light settings of 3000 K and 4000 K in this experiment, thus the data were analysed for each CCT separately. For each paired comparison, one score was obtained for the SPD that was perceived as more attractive, ranging from 1 to 3. For further analysis, the paired comparison scores were first transformed to a score for each of the two SPDs in the comparison, which is exactly the same procedure as for the Dutch study.¹⁷ The light setting selected as more attractive received a positive score, whereas the other setting in the same comparison received the same score but with a negative sign. As already mentioned, the four introductory SPD pairs were not included in the data analysis.

Analysis of variance (ANOVA) was performed per CCT, with SPD and application as fixed factors and the attractiveness scores as dependent variables. The ANOVA results indicate that for both CCTs, the SPD (3000 K: df = 6, F = 147.941, p < 0.001; 4000 K: df = 6, F = 42.911, p < 0.001) and the interaction between SPD and application (3000 K: df = 12, F = 7.010, p < 0.001; 4000 K: df = 12, F = 18.109, p < 0.001) have a significant influence on the perceived attractiveness. This means the influence of SPD on the object attractiveness varies between applications, as shown in Figure 5. The vertical axis represents the difference scores, relative to the SPD which has an R_a value of 100, i.e., the reference illuminant. Consequently, 3R100G100/4R100G100 get a score of 0, and the SPDs which were rated as more attractive than the reference received a positive score and the ones which were rated as less attractive received a negative score. Additional post hoc analysis per application revealed that the scores for some SPDs are not statistically significantly different from others at the 95% confidence level. These SPDs are encircled in Figure 5. For instance, in the left top graph of Figure 5, 3R80G110, 3R80G115 and 3R70G120 are encircled by a dotted line, which means there is no significant difference in attractiveness between these three SPDs.

From Figure 5, we can see that although under different CCTs, the preference order for fresh food and packaging material is almost the same. The SPD with highest R_a, i.e., R100G100, for both applications, is not the most preferred, nor R93G100, whose R_a value is the second highest. R70G120, R80G115 and R80G110, although their R_a values are only 70 or 80, are the top three most preferred SPDs. And there is no statistically significant difference in preference between these three SPDs. These SPDs all have relatively high G_a values (G_a ≥ 110), i.e., larger colour gamut areas. SPD R80G90, which also has an R_a value of 80 and a G_a of 90, is the least preferred SPD for both fresh food and packaging material.

Figure 5 also shows that user preference for white light sources is application dependent. The preference order for skin tone is quite different from the order for the other two applications. And the preference order is also different for different CCTs. Light settings 3R70G120 and 3R80G115, which are amongst the top three most preferred SPDs for the other two applications at 3000 K, are amongst the least preferred SPDs for skin tone and are significantly less attractive than the reference light source (3R100G100). The other five 3000 K light settings are not significantly different from each other. Moreover, 3R80G90 and 3R80G100, which is the first and second least preferred SPD for fresh food, obtained almost the same preference as 3R100G100. For 4000 K light settings, 4R80G90 and 4R80G100 are the first and second least preferred SPD, which is the same as the results for fresh food and packaging material.

For both CCTs, no gender difference was found with the Chinese people for any of the application areas. This is different from the findings obtained in the Dutch study, where a significant difference between genders was found.

4. Discussion

4.1. Comparison of user preference between the Dutch and Chinese studies

Comparing the results obtained from the Dutch study and Chinese study at 3000 K, it is found that the rank orders of SPDs for fresh food and packaging material are very similar, as shown in Figure 6. For both applications, Dutch people, the same as Chinese people, also prefer the light settings with an increased colour gamut size, i.e., 3R70G120, 3R80G115 and 3R80G110. The light setting with highest R_a value, i.e., 3R100G100, is not the most preferred SPD, and 3R80G90, whose G_a value is the lowest amongst all light settings, is always the least preferred one. However, for skin tone, Chinese people and Dutch people have a less clear but at least different preferences. 3R93G100 and 3R100G100 are amongst the most preferred light sources for Chinese people, whereas 3R70G120 and 3R80G115 are amongst the least preferred SPDs. For Dutch people, the latter two SPDs were amongst the most preferred light settings for skin tone rendering. There is a slight difference in the use of the scale, as shown in Figure 7, where the Dutch participants somewhat more strongly express their preference compared to the Chinese participants. This, however, cannot explain the differences in preference order for skin tone.

4.2. Relationship between user preference and colour fidelity (R_a), colour saturation (G_a)

The obtained results reveal that light settings with high colour fidelity are not amongst the most preferred by both Chinese and Dutch participants. Light settings with a larger colour gamut, i.e., higher G_a value, are preferred despite their reduced colour fidelity. Correlation analysis also shows that the difference scores correlate well with G_a for fresh food and packaging material (see Table 3).

Table 3

Correlation coefficients between the mean difference score and colour rendering index R_a, colour saturation index G_a (3000K NL: Dutch study with 3000K light settings; 3000K CN: Chinese study with 3000K light settings; 4000K CN: Chinese study with 4000K light settings).

	Difference score
	Fresh food			Packaging material			Skin tone
	3000 K NL	3000 K CN	4000 K CN	3000 K NL	3000 K CN	4000 K CN	3000 K NL	3000 K CN	4000 K CN
R _a	−0.184	−0.078	−0.157	−0.118	−0.204	−0.183	−0.104	0.911^a	0.267
G _a	0.919^a	0.951^a	0.963^a	0.891^a	0.956^a	0.971^a	0.461	−0.523	0.751

Correlation is significant at the 0.01 level (2-tailed).

As shown in Table 3, the Pearson correlation coefficients between the mean difference scores and G_a values are all higher than 0.89 for fresh food and packaging material in both Dutch and Chinese studies, which means the difference score has a strong positive correlation with G_a. As expected, the correlation between the mean difference score and R_a is very weak for both applications. For the skin tone application, the results are quite different. The correlation coefficient between mean difference score and R_a is 0.911 for 3000 K light settings in Chinese study, which implies that the attractiveness of skin has a positive correlation with colour fidelity (R_a). Increasing or decreasing in colourfulness would both lead to decreases in the attractiveness of human skin. We should, however, be very careful generalising this conclusion, because the difference scores for most 3000 K light settings are not significantly different from each other, as shown in Figure 6 (left bottom graph). The difference between the minimum and the maximum is only about 0.5. For the 4000 K light settings in the Chinese study and the 3000 K light settings in the Dutch study, the correlation between difference score is weak with both R_a and G_a. It should be noted that although the correlations are generally low for skin tone rendering, this does not mean there are no significant differences in terms of attractiveness between SPDs, as shown in Figure 8.

The scatter plots in Figure 8 clearly indicate that difference scores do not correlate well with R_a, but increase with G_a for fresh food and packaging material. Furthermore, we found that there is no significant difference in object attractiveness between the SPDs with G_a values of 110, 115 and 120, which is reflected in Figure 8 with the flattening of difference scores for G_a > 110. Due to the limited stimulus set, we do not know how attractiveness is rated when the colour gamut is further enlarged, i.e., G_a > 125. Ohno et al.¹¹ found that user preference would decrease when colours become too saturated. Hence, we anticipate that user preference would decrease when G_a is further increased for fresh food and packaging material. For skin tone in the Chinese study, user preference already starts decreasing when G_a is around 100 for 3000 K and 110 for 4000 K light settings.

Based on the expected decline in perceived attractiveness at high G_a values, we used second-order polynomial functions to develop models for predicting the user preference. For the Chinese study, the data of 3000 K and 4000 K for applications fresh food and packaging material are highly similar (Figure 8) and are therefore combined. The resulting equations (6) to (12) were derived with D_a indicating the difference in attractiveness compared to the reference illuminant R100G100. Their fit is indicated in Figure 9.

For fresh food,

\begin{matrix} Chinese D_{a} = - 2.17 \times 1 0^{- 3} G_{a}^{2} + 5.40 \times 1 0^{- 1} G_{a} \\ - 3.25 \times 1 0^{1} R^{2} = 0.963 \end{matrix}

(6)

\begin{matrix} Dutch D_{a} = - 3.19 \times 1 0^{- 3} G_{a}^{2} + 7.78 \times 1 0^{- 1} G_{a} \\ - 4.66 \times 1 0^{1} R^{2} = 0.901 \end{matrix}

(7)

For packaging material,

\begin{matrix} Chinese D_{a} = - 1.48 \times 1 0^{- 3} G_{a}^{2} + 3.73 \times 1 0^{- 1} G_{a} \\ - 2.27 \times 1 0^{1} R^{2} = 0.963 \end{matrix}

(8)

\begin{matrix} Dutch D_{a} = - 3.49 \times 1 0^{- 3} G_{a}^{2} + 8.34 \times 1 0^{- 1} G_{a} \\ - 4.93 \times 1 0^{1} R^{2} = 0.866 \end{matrix}

(9)

For skin tone,

\begin{matrix} Chinese 3000 K : \\ D_{a} = - 2.85 \times 1 0^{- 3} G_{a}^{2} + 5.84 \times 1 0^{- 1} G_{a} \\ - 3.00 \times 1 0^{1} R^{2} = 0.913 \end{matrix}

(10)

\begin{matrix} 4000 K : D_{a} = - 3.08 \times 1 0^{- 3} G_{a}^{2} + 6.88 \times 1 0^{- 1} G_{a} \\ - 3.83 \times 1 0^{1} R^{2} = 0.921 \end{matrix}

(11)

\begin{matrix} Dutch 3000 K : \\ D_{a} = - 3.25 \times 1 0^{- 3} G_{a}^{2} + 7.20 \times 1 0^{- 1} G_{a} \\ - 3.93 \times 1 0^{1} R^{2} = 0.414 \end{matrix}

(12)

The equations are different depending on the application and culture. For fresh food and packaging material, the curvature of the fit function for the Dutch study is larger than that for the Chinese study, as shown in the top and middle graphs of Figure 9. Larger curvature means that the same change in G_a value would lead to more difference in object attractiveness, which is in agreement with Figure 7. For skin tone, the optimum G_a value for 3000 K (G_a around 102) is smaller than that of 4000 K (G_a around 112) in the Chinese study, as shown in the bottom graphs in Figure 9. Furthermore, enhancing or decreasing colourfulness is not preferred for 3000 K light settings. In addition, the coefficient of determination (R²) for skin tone in the Dutch study, i.e., R²= 0.414, is not as high as the others, which means the equation does not accurately predict attractiveness. The bottom graph in Figure 9 also shows that the fit function for the 3000 K light setting in the Dutch study (black line) does not match the difference scores well.

Finally, we need to indicate that we have not systematically varied the colour gamut in all directions which has also been found to be important when assessing object attractiveness.²⁵ Therefore, it is difficult to determine the general applicability of the obtained equations, which should better be first validated with a larger set of preference experiments.

4.3. The influence of colour rendering indices (R_i) and relative chroma changes (C_i) on object attractiveness

In this section, the impact of colour rendering indices (R_i) and relative chroma changes (C_i) of TCS9 to TCS12, TCS13 and TCS15 will be presented.

The first eight CIE TCS are all moderately saturated colours, whereas TCS9 to TCS12 represent strong red, yellow, green and blue, respectively. TCS13 and TCS15 represent skin tones, with TCS13 for Caucasian skin tone and TCS15 for East-Asian skin tone.

Table 4 lists the Pearson correlation coefficients between the mean difference score and the colour rendering indices (R_i) and relative chroma changes (C_i) for the special TCS. The table shows that for fresh food and packaging material, the difference score does not correlate well with R values but has strong positive correlation with C values, especially C₉ (strong red). The correlation coefficient for C₉ is even higher than that for G_a except for the 3000 K light settings for packaging material in the Chinese study, which is 0.949, just a little lower than that for G_a (0.956). The correlation with C₁₀ (strong yellow) and C₁₁ (strong green) is not as strong as that with C₉, but still significant. Whereas, the correlation with C₁₂ (strong blue) is not significant for all situations. Hence, it implies that enhancing objects’ colour saturation, especially the red colour, would increase the user preference for a specific white light source. In addition, Table 4 shows that the correlation coefficients for C₁₃ and C₁₅ are also very high for fresh food and packaging material. TCS13 and TCS15 are the test colour samples representing skin tones, and both colour samples contain some red component. C₁₃ and C₁₅ show a high correlation with C₉ for the light settings in both the Chinese and Dutch studies (≥0.937, the correlation coefficient between C₉ and C₁₅ is even higher than 0.995). Therefore, it is not surprising that the mean difference score also correlates well with C₁₃ and C₁₅.

Table 4

Correlation coefficients between the mean difference score and colour rendering index and chroma change of some specific test-colour samples.

	Difference score
	Fresh food			Packaging material			Skin tone
	3000 K NL	3000 K CN	4000 K CN	3000 K NL	3000 K CN	4000 K CN	3000 K NL	3000 K CN	4000 K CN
R _a	−0.184	−0.078	−0.157	−0.118	−0.204	−0.183	−0.104	0.911^b	0.267
G _a	0.919^b	0.951^b	0.963^b	0.891^b	0.956^b	0.971^b	0.461	−0.523	0.751
R₉	0.344	0.433	0.244	0.373	0.315	0.216	0.337	0.666	0.650
R₁₀	0.183	0.542	0.548	0.305	0.362	0.553	−0.421	0.227	0.520
R₁₁	−0.715	−0.625	−0.685	−0.639	−0.694	−0.692	−0.519	0.884^b	−0.334
R₁₂	0.374	0.566	0.525	0.485	0.387	0.522	−0.361	0.118	0.463
R₁₃	−0.306	−0.179	−0.211	−0.224	−0.299	−0.227	−0.263	0.932^b	0.215
R₁₅	0.240	0.367	0.169	0.271	0.259	0.137	0.296	0.692	0.589
C ₉	0.980^b	0.974^b	0.995^b	0.975^b	0.949^b	0.995^b	0.405	−0.413	0.848^a
C ₁₀	0.866^a	0.813^a	0.853^a	0.855^a	0.755^a	0.848^a	0.197	−0.643	0.599
C ₁₁	0.826^a	0.861^a	0.863^a	0.798^a	0.834^a	0.874^a	0.294	−0.647	0.576
C ₁₂	0.790^a	0.808^a	0.136	0.837^a	0.748	0.087	−0.080	−0.522	0.286
C ₁₃	0.923^b	0.918^b	0.933^b	0.911^b	0.889^b	0.937^b	0.292	−0.575	0.709
C ₁₅	0.972^b	0.972^b	0.992^b	0.965^b	0.959^b	0.994^b	0.394	−0.434	0.834^a

Correlation is significant at the 0.05 level (2-tailed).

Correlation is significant at the 0.01 level (2-tailed).

For skin tone application, the correlation analysis shows different results for the 3000 K and 4000 K light settings as well as for different cultures. TCS15 is supposed to be used for characterising the colour rendering ability for East-Asian skin tones. From Table 4, we can see for the 4000 K light settings, the mean difference score has a significantly positive correlation with C₁₅ but not significantly with R₁₅, which implies that enhancing the skin colour could increase the attractiveness. However, the correlation coefficient for C₉ is also quite high, which implies that an increase in red is also appreciated for skin tone rendering at 4000 K. However, for 3000 K light settings in the Chinese study, the correlations with all C values are not significant although the correlation coefficients for R₁₃ (for Caucasian skin tone) and R₁₁ (strong green) are high: 0.932 and 0.884, respectively. For the Dutch study, not only R₁₃ and C₁₃, but also none of the other parameters listed in Table 4 can predict preference for skin tone rendering. Hence, it is difficult to give a clear conclusion for skin tone rendering from our study and further research is needed.

Because C₉ shows the highest correlation coefficients in Table 4, it is used to determine the parameters of the second-order polynomial fit functions for predicting the difference scores per culture, separately. The same data clustering was applied as for obtaining the equations for G_a (see Section 4.2).

For fresh food,

\begin{matrix} Chinese D_{a} = - 5.03 \times 1 0^{- 4} C_{9}^{2} + 5.75 \times 1 0^{- 2} C_{9} \\ + 1.19 \times 1 0^{- 1} R^{2} = 0.982 \end{matrix}

(13)

\begin{matrix} Dutch D & _{a} = - 1.04 \times 1 0^{- 3} C_{9}^{2} + 6.86 \\ \times 1 0^{- 2} C_{9} - 3.81 \times 1 0^{- 2} \\ R^{2} = 0.987 \end{matrix}

(14)

For packaging material,

\begin{matrix} Chinese D_{a} = - 3.12 \times 1 0^{- 4} C_{9}^{2} + 4.29 \times 1 0^{- 2} C_{9} \\ + 1.86 \times 1 0^{- 1} R^{2} = 0.950 \end{matrix}

(15)

\begin{matrix} Dutch D_{a} = - 1.08 \times 1 0^{- 3} C_{9}^{2} + 6.51 \times 1 0^{- 2} C_{9} \\ - 1.25 \times 1 0^{- 1} R^{2} = 0.980 \end{matrix}

(16)

For skin tone,

\begin{matrix} Chinese 3000 K : \\ D_{a} = - 1.69 \times 1 0^{- 3} C_{9}^{2} - 1.67 \times 1 0^{- 2} C_{9} \\ - 4.65 \times 1 0^{- 2} R^{2} = 0.878 \end{matrix}

(17)

\begin{matrix} Chinese 4000 K : \\ D_{a} = - 1.36 \times 1 0^{- 3} C_{9}^{2} + 2.32 \times 1 0^{- 2} C_{9} \\ - 3.17 \times 1 0^{- 2} R^{2} = 0.986 \end{matrix}

(18)

\begin{matrix} Dutch 3000 K : \\ D_{a} = - 1.55 \times 1 0^{- 3} C_{9}^{2} + 9.66 \times 1 0^{- 3} C_{9} \\ & + 4.79 \times 1 0^{- 1} R^{2} = 0.365 \end{matrix}

(19)

The coefficients of determination, R², for C₉ are mostly a little higher than those for G_a, especially for the Dutch study. Hence, both G_a and C₉ are good indicators for predicting user preference for white-light sources, except for skin tone rendering in the Dutch study. Figure 10 shows an increase in attractiveness for fresh food and packaging material as a function of C₉, even above 20. As explained earlier, we did not systematically vary C₉ and we did not include values above 25. Therefore, to determine the optimal C₉ value, further study is needed.

Nowadays, R₉ is sometimes provided in addition to R_a for indicating the colour rendering index for the strong red colour, where, similar to R_a, a higher R₉ is often assumed to be better. However, our findings reveal that user preference does not correlate well with R₉. It only represents the similarity of red objects rendered by the test light source compared with the reference illuminant. So, enhancing or weakening the red appearance of objects would both lead to a reduction of the R₉ value. Hence, R₉ appears not to be a good predictor for rating the perceived attractiveness of familiar objects.

4.4. Summarising the Dutch and Chinese studies

From both Dutch and Chinese studies, we found that neither R_a nor R₉ can correctly predict the preference order for a set of white-light sources, in terms of object attractiveness. The light setting with the highest colour fidelity (R_a ≈100) was not amongst the most preferred ones. Light settings with R_a values of ≈80, even 70, in combination with a high G_a value, are preferred over the reference light source with an R_a of 100. For example, R80G115 is amongst the most preferred light settings, but R80G90 is the least preferred one. Other studies^25,30 have found similar results, i.e., colour rendering indices on their own cannot predict people’s colour preference.

The CSI (G_a) and relative chroma change for the strong red TCS (C₉) are both found to be good predictors for rating object attractiveness. User preference has a strong positive correlation with both G_a and C₉. Enhancing object colour, in particular for the strong red TCS, increases people’s appreciation towards a specific white-light source, but is in current specifications penalised by lower R_a and R₉ values. The top three most preferred SPDs in our study, all have relatively larger gamut areas (G_a ≥ 110), and their red saturation was enhanced (C₉ > 9, C₉ of 4R70G120 is even as high as 20). While the SPD R80G90 (G_a = 90, C₉ = −21 for the 3000 K light setting and −26 for the 4000 K light setting) was the least preferred SPD amongst all the seven test light settings. Jost-Boissard et al.^12,13 also found there was high correlation between attractiveness and gamut area based indices,¹³ and fruits and vegetables were found to be rated more attractive when their colours appeared more saturated¹² in their study. Royer et al.²⁵ concluded that increased saturation of reddish colours was more preferred, whereas increased saturation of yellow was less appreciated. Wei et al.³⁰ also found with their study that the preference of the light settings with chroma enhancement in blue and yellow was lower than those settings with enhancement in red colour. In our study, we also found that the user preference does not correlate well with C₁₂ (blue). However, the correlation with C₁₀ (yellow) is again significant and positive. According to Tables 1 and 2, it appears that C₁₀ and C₁₂ do not cover a sufficiently wide range for drawing conclusions. Hence, combining our own findings and the literature, it seems that enhancing red colour is important for improving the user appreciation for white-light sources. According to Ohno et al.,¹¹ the increase in colourfulness should not be too high otherwise the preference starts decreasing. This finding was also anticipated with the second-order polynomial fits in our study, although we did not find a decrease in attractiveness for the applications fresh food and packaging material. Perhaps material, perhaps because the increase in colour gamut and red chroma was not large enough. An increase in colour saturation always goes hand in hand with a decrease in colour fidelity. Since colour fidelity is still one important aspect for colour rendition, there inevitably is a trade-off between increasing colour saturation and maximising colour fidelity.

Other than colour saturation, user preference for white light sources was also affected by some other factors, like application, culture, and CCT. The preference order for fresh food and packaging material is almost equal despite different cultures (Dutch and Chinese) and different CCTs (3000 K and 4000 K), but quite different from the order for skin tone. In contrast to the other two applications, red enhancement was not appreciated by Chinese people, for 3000 K, because their skin appeared too reddish. The red enhanced SPD, 3R70G120 (C₉ = 15), was the least preferred one amongst the 3000 K light settings. However, this is not true for the more bluish 4000 K light settings. 4R70G120 (C₉ = 20) is amongst the most preferred SPDs amongst the 4000 K light settings. At 4000 K, chroma change C₁₅ (TCS15 is for East-Asian skin tones) seems to be a good predictor for rating the attractiveness of skin tone. Enhancing skin colour could increase the attractiveness of human skin, but for the 3000 K light settings we have not found this effect. The user preference for skin tone is also different for different cultures. In the Dutch study, the red enhanced SPDs, 3R70G120 (C₉ = 17.5) and 3R80G115 (C₉ = 10), are amongst the top three most preferred SPDs. But another red enhanced SPD, 3R80G110 (C₉ = 13.5), was rated as the second least preferred light setting. TCS13 is supposed to be used for characterising the colour rendering ability for Caucasian skin tones. But in the Dutch study, neither R₁₃ nor C₁₃ was found to correlate well with the preference scores. Moreover, the correlation between the user preference and all the parameters, including R_a, G_a, R_i and C_i, is found to be not significant. Compared with fresh food and packaging material, people have a less clear preference and it is not possible to draw firm conclusions from this study to identify optimal light settings for skin tone rendering and further research is necessary.

5. Conclusions

The widely used CIE general colour rendering index (R_a) characterises the colour fidelity of a white-light source. It cannot be used for predicting the attractiveness of rendered object colours, i.e., a light source with a higher R_a value is not necessarily more preferred. In addition, the supplementary special colour rendering index for the strong red TCS, R₉, nor any of the other special colour rendering indices show a clear relation with user preference.

The colour-gamut-based CSI (G_a) and chroma change for the strong red TCS (C₉), on the other hand, can be used for predicting changes in object attractiveness for fresh food and packaging material, albeit with different fit functions per application and culture. We can conclude that if the colour fidelity is high enough (e.g. ≥ 70), increasing red saturation has a positive contribution to the appreciation of a white-light source. The second-order polynomial fit functions suggest a decrease in attractiveness when C₉ becomes too large, but the range of C₉ values used in our study was not sufficient to confirm this.

For skin tone rendering, Chinese participants did not like an increase, nor a decrease, in red saturation for the 3000 K light settings, but they allowed an increase in red saturation for the 4000 K light settings. The attractiveness for skin tone rendering could also be predicted with second-order polynomial functions for the Chinese study. There was no clear relation between objective measures and skin tone preference for the Dutch study.

We have found that object attractiveness is application-dependent, gender-specific, culturally related and is also influenced by the correlated colour temperature of the white-light source. So, we cannot clearly identify specific requirements for optimal attractiveness in the two-dimensional R_a–G_a space, nor for specific C₉ values. Moreover, the equations to predict changes in attractiveness from the C₉ or G_a values are different per application and culture and are based on a limited set of light settings. Nonetheless, for providing a more complete description for the evaluation and specification of the colour rendering properties of white-light sources, we propose using the CIE general colour rendering index (R_a) in conjunction with the CSI (G_a) and the relative chroma change for the strong red TCS (C₉).

Footnotes

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors are employed by Philips Lighting.

References

Commission Internationale de l’Eclairage. International Lighting Vocabulary 17-221 Colour rendering (of a light source). Vienna: CIE, 2011.

Commission Internationale de l’Eclairage. International Lighting Vocabulary 17-154 CIE 1974 General Colour Rendering Index [Ra]. Vienna: CIE, 2011.

Commission Internationale de l’Eclairage. Method of Measuring and Specifying Colour Rendering Properties of Light Sources. CIE Publication 13.3. Vienna: CIE, 1995.

Houser

Mossman

Smet

Whitehead

. Colour rendering and its applications in lighting. Leukos 2016; 12: 7–26.

Commission Internationale de l’Eclairage. Colour Fidelity Index for Accurate Scientific Use. CIE 224: 2017. Vienna: CIE, 2017.

Illuminating Engineering Society of North America. IES Method for Evaluating Light Source Color Rendition, TM-30-15. New York: IESNA, 2015.

David

Fini

Houser

Ohno

Royer

Smet

KAG

Wei

Whitehead

. Development of the IES method for evaluating the color rendition of light sources. Optics Express 2015; 23: 15888–15906.

Royer

. IES TM-30-15 is approved – Now what? Moving forward with new color rendition measures. Leukos 2016; 12: 3–5.

Smet

Ryckaert

Pointer

Deconinck

Hanselaer

. Correlation between colour quality metric predictions and visual appreciation of light sources. Optics Express 2011; 19: 8151–8166.

10.

Houser

Wei

David

Krames

Shen

. Review of measures for light-source colour rendition and considerations for a two-measure system for characterizing colour rendition. Optics Express 2013; 21: 10393–10411.

11.

Ohno Y, Fein M, Miller C. Vision experiment on chroma saturation for colour quality preference: Proceedings of 28th Session of the CIE, Manchester, UK, June 28 to July 4: 2015: 60–69.

12.

Jost-Boissard

Fontoynont

Blanc-Gonnet

. Perceived lighting quality of LED sources for the presentation of fruit and vegetables. Journal of Modern Optics 2009; 56: 1420–1432.

13.

Jost-Boissard

Avouac

Fontoynont

. Assessing the colour quality of LED sources: Naturalness, attractiveness, colourfulness and colour difference. Lighting Research and Technology 2015; 47: 769–794.

14.

Wei

Houser

Allen

Beers

. Colour preference under LEDs with diminished yellow emission. Leukos 2014; 10: 119–131.

15.

Wei

Houser

David

Krames

. Perceptual responses to LED illumination with colour rendering indices of 85 and 97. Lighting Research and Technology 2015; 47: 810–827.

16.

Teunissen

van der Heijden

FHFW

Poort

SHM

de Beer

. Characterising user preference for white LED light sources with CIE colour rendering index combined with a relative gamut area index. Lighting Research and Technology 2017; 49: 461–480.

17.

Teunissen

Hoelen

. Progress in characterizing the multidimensional colour quality properties of white LED light sources. Proceedings of the SPIE 2016; 9768: 1–12. . DOI: 10.1117/12.2217661.

18.

Guo

Houser

. A review of colour rendering indices and their application to commercial light sources. Lighting Research and Technology 2004; 36: 183–199.

19.

Rea

Freyssinier-Nova

. Colour Rendering: A tale of two metrics. Colour Research and Application 2008; 33: 192–202.

20.

Freyssinier

Rea

. A two-metric Proposal to Specify the Colour-Rendering Properties of Light Sources for Retail Lighting: Proceeding of SPIE 2010, Tenth International Conference on Solid State Lighting,, San Diego, USA. August 2–4: 2010: 7784: 77 840V–1.

21.

Houser

. If not CRI, then what?. Leukos 2013; 9: 151–153.

22.

Khanh

Bodrogi

Vinh

Stojanovic

. Colour preference, naturalness, vividness and colour quality metrics, Part 1: Experiments in a room. Lighting Research and Technology 2017; 49: 697–713.

23.

Khanh

Bodrogi

Vinh

Stojanovic

. Colour preference, naturalness, vividness and colour quality metrics, Part 2: Experiments in a viewing booth and analysis of the combined dataset. Lighting Research and Technology 2017; 49: 714–726.

24.

Khanh TQ, Bodrogi P. Colour preference, naturalness, vividness and colour quality metrics, Part 3: Experiments with makeup products and analysis of the complete warm white dataset. Lighting Research and Technology First published 16 September 2016. DOI: 1477153516669558.

25.

Royer

Wilkerson

Wei

Houser

Davis

. Human perceptions of colour rendition vary with average fidelity, average gamut, and gamut shape. Lighting Research and Technology 2017; 49: 992–1014.

26.

Japanese Standards Association, JIS Z 8726-1990 Method of Specifying Colour Rendering Properties of Light Sources. Tokyo: Japanese Standards Association, 1990.

27.

Commission Internationale de l’Eclairage. ISO 11664-5:2016(E). Colorimetry Part 5: CIE 1976 L*u*v* Colour Space and u’, v’ Uniform Chromaticity Scale Diagram. Vienna: CIE, 2016.

28.

China National Standard, GB11533-2011

(Standard vision logarithm tables). Available from http://www.spsp.gov.cn/page/P1306/519.shtml.

29.

Cao

. Colour Blind Examination (Version 5), Beijing: Peoples Health Press, pp. 2016–2016.

30.

Wei

Houser

David

Krames

. Colour gamut size and shape influence colour preference. Lighting Research and Technology 2017; 49: 992–1014.

The appreciation of LED-based white light sources by Dutch and Chinese people in three application areas

Abstract

1. Background

1.1. Limitations of colour fidelity indices

1.2. A simple CRI-based colour gamut index

1.3. CIELUV-based relative chroma changes

1.4. Aim of the current study

2. Colour rendering preference study for LED-based white-light sources in China

2.1. Set-up

2.2. Method

2.3. Participants

3. Results for the Chinese study

4. Discussion

4.1. Comparison of user preference between the Dutch and Chinese studies

4.2. Relationship between user preference and colour fidelity (Ra), colour saturation (Ga)

4.3. The influence of colour rendering indices (Ri) and relative chroma changes (Ci) on object attractiveness

4.4. Summarising the Dutch and Chinese studies

5. Conclusions

Footnotes

Declaration of conflicting interests

Funding

References

4.2. Relationship between user preference and colour fidelity (R_a), colour saturation (G_a)

4.3. The influence of colour rendering indices (R_i) and relative chroma changes (C_i) on object attractiveness