Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Contents lists available at ScienceDirect
Optik journal homepage: www.elsevier.com/locate/ijleo
Original research article
Optimising colour preference and colour discrimination for jeans under 5500 K light sources with different Duv values Ying Liua, Qiang Liua,b,*, Zheng Huanga, Michael R. Pointerc, Lianjiang Raod, Zhen Houa a
School of Printing and Packaging, Wuhan University, Wuhan, 430079, China Hubei Engineering Research Center for Digitization and Virtual Reproduction of Colour Information in Cultural Relics, Wuhan, 430079, China c School of Design, University of Leeds, Leeds, LS2 9JT, UK d WAC Lighting, Shanghai, 201204, China b
A R T IC LE I N F O
ABS TRA CT
Keywords: Colour preference Colour discrimination Blue jeans Farnsworth-Munsell 100 Hue Test
Our previous work revealed that, compared with other correlated colour temperatures (CCTs), a light source of 5500 K could arouse the strongest colour preference perception for blue jeans and provide the best colour discrimination capability for blue colours. In this study, we further investigated the impact of light sources on colour preference and colour discrimination using the same experimental objects (i.e. jeans and colour samples). Nine light sources of 5500 K were adopted and the illuminance level was set to 500 lx. Those lights were of different Duv values (-0.02 to 0.02, in 0.005 intervals) and similar colour rendering indices (CRIs, 87–92). Following a similar experimental protocol adopted in the earlier work, 30 subjects participated in the colour preference experiment while 24 observers joined the colour discrimination test (the FarnsworthMunsell 100 Hue Test). The experimental results indicate that, once again, the colour preference and colour discrimination of lighting reached an optimum simultaneously, at a Duv value of -0.01. The hue shift and hue difference were proved to be closely associated with colour preference rating and colour discrimination scores, respectively. In addition, significant gender difference was again found, which had been reported in our latest work.
1. Introduction Colour preference [1–6] and colour discrimination [7–11] are two of the most important dimensions for evaluating the colour quality of lighting. Colour preference of lighting refers to the observers’ visual colour appreciation under different light sources. This issue is usually investigated by psychophysical studies, with various light sources or experimental objects as independent variables and human preference ratings as dependent variables. The aim of these studies are to investigate under which kind of light sources subjects prefer the rendered colours of the illuminated object [5,12–14], to explore the influencing factors of visual colour preference perception [1–3,12,14–19] and to establish an objective metric for predicting the subjective responses of the observers [3,4,20–26]. As for colour discrimination of lighting, it is defined as the capability of light sources which “allows the observer to discriminate among a large variety of object colors simultaneously viewed” [11]. In the current literature, this capability is usually quantified by use of the Farnsworth-Munsell (FM) 100 Hue Test [27–30]. In that test, observers are asked to sort 85 moveable samples with perceptually uniform hue steps into a gradient order and their discrimination ability is then represented by an error score computed from their
⁎
Corresponding author at: School of Printing and Packaging, Wuhan University, Wuhan, 430079, China. E-mail address:
[email protected] (Q. Liu).
https://doi.org/10.1016/j.ijleo.2019.163916 Received 25 September 2019; Accepted 26 November 2019 0030-4026/ © 2019 Elsevier GmbH. All rights reserved.
Please cite this article as: Ying Liu, et al., Optik - International Journal for Light and Electron Optics, https://doi.org/10.1016/j.ijleo.2019.163916
Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Y. Liu, et al.
sorted orders and the standard order. The colour discrimination ability of a light source can be obtained by averaging the error scores of each subjects. In the current stage, most researchers investigating the colour preference of lighting asked the subjects to observe the colour appearance of the experimental objects under different light sources and then respond with their preference ratings. The experimental objects were commonly of multiple colour, including artworks [31], consumer goods [32,33], fruit and vegetables [34,35], printed images [36], makeup products [37], as well as combined objects [38,39]. However, to our knowledge, few studies have been conducted which focus on the colour preference of single-colour objects with a colour gradient pattern. In addition to this, the past studies investigated the topics of colour preference and colour discrimination separately, leaving the correlation between the two dimensions unclear. Therefore, in our recent work [30], nine lights of different CCTs (2500 K–6500 K, 500 K interval, 200 lx) were used to investigate the colour preference of 7 pairs of blue jeans. Meanwhile, five of those lights (2500 K–6500 K, 1000 K interval) were used to test the influence of lighting on distinguishing blue colours by the FM-100 test. The main finding of that work was that both colour preference and colour discrimination reach optimum at a CCT of 5500 K, which indicates an optimal solution for display lighting for jeans—under such a CCT, retailers could make their jeans more attractive in colour while consumers will discriminate more colours and thus be able to make better choices. Based on the conclusion of our latest work, in this study the impact of different Duv values (the distance from the test chromaticity coordinates to the Planckian locus) on colour preference and colour discrimination was further investigated. Nine 5500 K light sources of constant CRI but different Duv values were adopted. The colour preference experiment (9 lights, Duv ranges from -0.02 to 0.02 with 0.005 interval) and colour discrimination test (5 lights, Duv ranges from -0.02 to 0.02 with 0.01 interval) were carried out following similar protocols to those of our previous work [30], with the same experimental objects. Note that in this study the illuminance level was set to 500 lx for a better simulation of the actual lighting condition in retail outlets and shopping malls. In addition, unlike our former work which only implemented the FM-100 Hue test with its Green-Blue and Blue-Purple boxes, in this study all of the 4 boxes (Red-Green, Green-Blue, Blue-Purple, Purple-Red) were used for quantifying the colour discrimination capability of the experimental light sources. 2. Method 2.1. Experimental setup The experimental setup of this work is similar with that of our previous study [30]. Both the colour preference experiment and the colour discrimination test were implemented in a light booth with a size of 50 cm × 50 cm × 60 cm (W × D×H). The walls and floor of that booth were uniformly painted with Munsell N7 matt grey paint. A chair was placed 40 cm in front of the booth. During the test, the participant was asked to adjust the height of the chair, so that the illuminant unit of the light booth would not be seen. Nine spectral power distributions (SPDs) were generated by a spectrally-tuneable lighting system (LEDcube, Changzhou Thouslite Ltd), as shown in Fig. 1. Those lights were of similar CCT (approximately 5500 K) and CRI (87–92) values, but their Duv values ranged from -0.02 to 0.02. The illuminance level around the floor of the booth was 500 ± 10 lx, which corresponds to a luminance value around 170 cd/m2. An X-Rite i1 Pro 2 spectrophotometer was used to calibrate those SPDs while the illuminance level was measured by a Testo 540 illuminance meter. The colorimetric parameters of the experimental lights are shown in Table 1, together
Fig. 1. Relative spectral power distributions of the experimental light sources. 2
Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Y. Liu, et al.
Table 1 The colorimetric properties of the experimental SPDs and the corresponding values of typical colour quality metric. ID*
#1
#2
#3
#4
#5
#6
#7
#8
#9
CCT(K) Duv u' v' CRI GAI FSCI Qa (9.0.3) Qf (9.0.3) Qg (9.0.3) Qp (7.4) FCI(CAM02) CDI CSA CPI CRI-CAM02UCS CRI2012 MCRI Rf Rg △C* CQI-1 CQI-2 GAI-RA GVI WS Sneutral DSI(D65) Percent tint
5484 0.020 0.192 0.498 89 75 87 82 82 91 81 96 110 0.055 139 81 87 83 82 92 −2.13 173 1782 82 92 0.26 2.08 88 0.07
5481 0.015 0.196 0.493 90 80 89 84 84 93 84 97 117 0.057 140 83 89 86 83 93 −1.49 178 1891 85 92 0.33 3.06 89 0.16
5591 0.010 0.198 0.487 89 86 91 86 85 96 86 96 126 0.059 139 85 91 87 85 96 −0.78 180 2053 88 92 0.43 4.31 90 0.32
5497 0.005 0.203 0.483 92 92 94 91 91 99 93 96 134 0.061 144 91 95 90 91 99 −0.11 185 2134 92 91 0.54 5.63 93 0.45
5500 −0.000 0.207 0.477 89 95 94 91 91 100 94 99 138 0.062 138 90 94 91 88 98 0.03 187 2160 92 89 0.69 6.84 92 0.56
5467 −0.005 0.211 0.473 88 99 95 93 92 102 97 101 145 0.063 137 92 95 92 90 100 0.55 189 2236 94 88 0.88 7.60 92 0.69
5518 −0.010 0.214 0.468 91 105 97 97 94 105 101 97 153 0.065 141 94 97 93 95 104 1.34 191 2389 98 88 1.10 7.91 94 0.87
5477 −0.015 0.218 0.463 90 107 97 95 93 105 99 96 156 0.066 140 92 96 93 93 104 1.33 190 2372 99 86 1.40 7.66 94 1.00
5576 −0.020 0.222 0.457 87 117 99 92 86 112 102 93 171 0.069 135 83 88 93 83 111 3.08 191 2704 102 86 1.85 6.84 89 1.30
* (u’, v’): CIE 1976 chromaticity coordinates, Duv: distance from the test chromaticity coordinates to the Planckian locus, CRI: CIE General Colour Rendering Index [40], GAI: Gamut Area Index [41], FSCI: Full Spectrum Colour Index [42], CQS: Colour Quality Scale (Qa, Qf, Qp, Qg) [43], FCI: Feeling of Contrast Index [44], CDI: Colour Discrimination Index [45], CSA: Cone Surface Area [46], CPI: Color Preference Index [47], CRICAM02UCS: Colour Rendering Index calculated in CAM02UCS [48], CRI2012: Updated version of CRI [49], MCRI: Memory Colour Rendering Index [50], Rf and Rg: IESNA TM-30 metrics [51], △C*: mean chroma shift of CQS [33,37,52], CQI-1 and CQI-2: two of the latest combined metrics named Colour Quality Index [33,52], GAI-RA: the arithmetic mean value of GAI and CRI [26,34], GVI: Gamut Volume Index [53], WS: White Sensation [54], Sneutral: Degree of neutrality [55], DSI: Daylight Spectrum Index [56], percent tint: a recently proposed whiteness metric for lighting [57].
with several typical colour quality measures. Such data are shown with the aim of helping following researchers to investigate the impact of different colour quality measures upon colour preference of lighting.
2.2. Experimental design During the colour preference experiment, the 7 pairs of blue jeans adopted in our previous work [30] were used again in this study. Those jeans were selected from a pool of 55 jeans and their colours located uniformly in CAM16UCS [58]. Thirty observers, 15 males and 15 females, were invited to participate in the preference rating test. These subjects were 18–28 years old, with an average age of 21.1 years. All of the observers had passed the Ishihara Colour Vision Test while none knew the research purpose before the experiment. During the preference test, the subjects were asked to respond with their visual colour preference for the experimental jeans, with the same 7-point rating method used before. That is, they were instructed to express strongly dislike, moderately dislike, slightly dislike, neutral, slightly like, moderately like and strongly like by scores of -3, -2, -1, 0, 1, 2, 3, respectively. Meanwhile, a randomly selected light source was exhibited twice in the preference rating section, with the aim of quantifying the intra-observer variability of each participant. For the colour discrimination test, the FM-100 Hue Test procedure was followed. Twenty-four subjects, 12 males and 12 females (17–32 years old, with a mean of 20.2 years), participated in this experiment. Note that in this test only five lights from Table 1 were used (i.e. Duv=-0.02, -0.01, 0, 0.01, 0.02), since adopting all the lights in this test would take too much time and lead to visual fatigue. Meanwhile, as stated earlier, in our former work only two boxes (Green-Blue, Blue-Purple) containing blue colour samples were used in the FM-100 Hue Test. In this study, however, all of the four boxes (Red-Green, Green-Blue, Blue-Purple, Purple-Red) were used for quantifying the colour discrimination capability of the experimental light sources. The aim of such a setting is to further investigate the impact of lighting upon various colour regions (some of those findings will be discussed in future papers).
3
Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Y. Liu, et al.
2.3. Experimental procedure In the colour preference experiment, the procedure is quite consistent with our previous work [30]. Upon arrival, the observer was sequentially asked to implement the Ishihara Colour Vision Test, to wear a grey coat and then to sign a consent form. After being escorted to the booth, the observer was asked to adjust the height of the chair to make sure that he/she could not see the illuminant unit of the booth. Then, the instructions for the test were presented orally by an experimenter and the room lighting in the lab was turned off, leaving the experimental light source the only illumination in the room. During the test, the observers responded orally with their subjective ratings. Before the formal experiment, the observer was given 20 s to visually adapt to a welcome light, which was randomly picked up from the nine experimental light sources. The experimenter then provided a training session in which the participant was asked to rate his/her whiteness perception of the lighting, with no object in the booth. The aim of such a setting was to help the observer become familiar with the 7-point rating approach. Then, the formal test began. The experimenter asked the observer to close their eyes when the experimental light was changed. This step lasted for 20 s and it aimed to eliminate the short-term memory effect of the former lighting (This step was repeated every time the experimental light was changed). When the observer opened his/her eyes, he/she was given as much time as they needed to fully adapt to the new light in the empty booth and then make his/her judgment. Note that during this test, the whiteness rating experiment was conducted first and carried out only once, with no object in the booth (These whiteness data were used to investigate the correlation between the whiteness of lighting and the colour preference of lighting, as reported in our recent papers [1,12]). As for colour preference rating, after the observer visually adapted to the light, the experimenter put one pair of jeans into the booth and then asked the observer to proceed with the colour preference rating. During this step, the experimenter reminded the observer to concentrate on the colour of jeans. After the subject finished the ratings and confirmed their scores, the experimenter changed the light source. This procedure was repeated for each of the experimental lights (including the repeated light source for intra-observer variability) and then for each pair of jeans, which means that observers had to rate with their colour preference perception of one pair of jeans under different lights and then repeat the trial for a second pair of jeans). In addition, the presentation order of the experimental jeans and the light sources was randomised and counterbalanced between the subjects. For the colour discrimination experiment, the same FM-100 Hue Test was carried out in the same booth with the same lights (five, as described above). The observers were asked to rearrange the disordered colour samples such that their hue attributes were distributed in a continuous manner. In this test, similarly, the presentation order of the experimental light sources, as well as the four boxes corresponding to different colour regions, was randomised and counterbalanced between observers. After one subject finished the whole test under a light, his/her colour discrimination capability under such illumination could be quantified by an error score, which was computed as the sum of the differences between the number of the colour sample and the numbers of the two samples placed adjacent to it. In addition, it should be mentioned that after the visual test, when we calculated the error scores, the huetransposition effect caused by the experimental light sources was considered. That is, only the colour sample transposition caused by the participants were used to calculate the error scores (defined as adjusted error scores), the transpositions caused by light sources were excluded. 3. Results and discussion The overall results of the colour preference and colour discrimination experiments are summarised in Tables 2 and 3, respectively. Note that in Table 3, the average adjusted error scores of the FM-100 Hue Test are presented thus a lower score corresponds to a better colour discrimination capability. In fact, in this study only the relative position of Chip #41 and Chip #42 were impacted by Table 2 The Avg value and SD of preference rating for different jeans under the lighting conditions with different Duv values, together with the Pearson correlation coefficient r between the Avg and the SD of each scenario. Object
Stat.
Duv = 0.020
Duv = 0.015
Duv = 0.01
Duv = 0.005
Duv = 0
Duv=-0.005
Duv=-0.01
Duv=-0.015
Duv=-0.02
r
Jean 1
AVG SD AVG SD AVG SD AVG SD AVG SD AVG SD AVG SD AVG
−1.47 1.15 −1.33 1.53 −1.43 1.56 −1.27 1.55 −1.13 1.52 −1.47 1.45 −1.23 1.52 −1.33
−0.97 1.38 −0.27 1.46 −1.00 1.24 −0.87 1.33 −1.03 1.45 −1.20 1.54 −0.60 1.58 −0.85
0.03 1.33 0.30 1.37 0.03 1.25 0.37 1.25 −0.20 1.38 −0.17 1.42 0.07 1.26 0.06
0.73 1.12 0.60 1.20 0.93 1.09 0.87 1.18 0.77 1.20 0.67 1.11 0.50 1.20 0.72
1.37 0.95 0.77 0.80 1.00 0.86 1.23 1.05 0.83 1.10 0.97 1.08 0.93 1.03 1.01
1.23 1.09 1.10 1.11 1.20 1.14 1.33 1.07 1.13 1.31 1.20 1.22 1.03 1.38 1.18
1.30 1.24 1.17 1.32 1.43 1.05 1.27 1.24 1.40 1.20 1.23 1.15 0.77 1.41 1.22
1.20 1.08 0.97 1.20 1.27 1.24 1.10 1.27 1.00 1.53 0.90 1.42 0.53 1.63 1.00
0.90 1.35 0.53 1.65 1.03 1.45 0.70 1.62 0.83 1.46 0.63 1.60 0.43 1.78 0.72
−0.44
Jean 2 Jean 3 Jean 4 Jean 5 Jean 6 Jean 7 Overall
The highest average preference rating is underlined and bolded. Avg: average; SD: standard deviation. 4
−0.53 −0.57 −0.60 −0.52 −0.59 −0.35 –
Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Y. Liu, et al.
Table 3 The Avg and SD of the adjusted error score of the Farnsworth-Munsell-100 Hue Test, together with the Pearson correlation coefficient r between the Avg and the SD. Box
Stat
Duv = 0.020
Duv = 0.010
Duv = 0.000
Duv=-0.010
Duv=-0.020
r
Red-Green
AVG SD AVG SD AVG SD AVG SD AVG
3.33 4.57 8.17 5.94 7.17 6.43 3.00 4.04 5.42
2.50 3.80 6.67 3.59 4.83 4.76 2.33 3.25 4.08
1.50 3.62 8.00 4.90 5.17 4.83 1.33 1.89 4.00
1.17 1.82 6.00 3.65 3.50 4.80 1.33 2.49 3.00
0.83 1.99 9.33 5.73 4.33 4.31 2.50 3.23 4.25
0.90
Green-Blue Blue-Purple Purple-Red Overall
0.90 0.87 0.96 –
The minimum average adjusted error score is underlined and bolded.
the experimental light sources, thus the findings based on adjusted error scores and original error scores are very consistent. From these two tables, it is clear that the ratings of colour preference and colour discrimination reach an optimum value simultaneously (especially for the blue regions), with a Duv value of -0.01. Detailed analysis will be presented in the following sections. 3.1. Inter-observer and intra-observer variability The inter-observer variability of the ratings for the colour preference experiment was denoted by the standard deviations of the preference scores, as shown in Table 2. It is obvious that those measures, with regard to each trial, are consistent. In addition, such results are also similar with those of our former study, in which the same experimental jeans were adopted but with different light sources used in the visual experiment [30]. As stated above, the intra-observer variability was verified by asking the observers to rate a randomly selected light source twice. Afterwards, the intra-observer variability was quantified by the absolute-difference method, which has been used in many of our studies [1,2,13,30]. According to that method, an abnormal data pair was defined as two responses for a same light whose absolute difference was larger than 2. Thus, the intra-observer variability was represented by the ratio of the number of abnormal data pairs to the total number of data points (excluding the repeated data). In this study, the average intra-observer variability for different jeans is 6.2 %, which is similar with that of our recent work with the same jeans (4.8 %) [30]. 3.2. Analysis of colour preference It can be seen from Table 2 that the preference ratings for light sources with negative Duv values are significantly higher than those of lights with positive Duv values and the rating trends of different jeans under different light sources are similar. Such rating consistency highlights the dominate effect of lighting on colour preference perception and validates our initial opinion to find the best light source for jeans. This statement is strengthened by a repeated measures analysis of variance (rm-ANOVA), according to which the impact of light sources in colour preference ratings is significant (F = 143.92, p < 0.001) while the impact of the experimental jeans is not (F = 2.52, p = 0.803). Meanwhile, it is quite interesting that the impact of gender (F = 22.50, p < 0.001), as well as the interaction of gender and lights (F = 13.81, p < 0.001) are both significant: detailed results will be described below. According to the work of Wei and Houser, the negative Duv values were appreciated more since they generally exhibited higher scores for relative gamut and colour fidelity [5]. In our latest work, however, it has been demonstrated that such a result is also closely associated with the human visual preference for the whiteness of lighting [1,12]. In addition, in Table 2 it can been seen that the average rating and standard deviation of each trial correlated negatively (r < 0). This finding is well correlated with several past studies [2,3,13,30,36,59], which indicates that when subjects appreciate certain lights they will respond consistently while if they do not like certain lights their responses will be relatively diverse. As for colour appearance analysis, in our latest work we found that, in CAM16USC space, the chroma and hue of the experimental jeans respectively decreased and increased with CCTs while the preference ratings increased first with CCT from 2500 K to 5500 K and then decreased from 5500 K to 6500 K [30]. In that work, we speculated that the hue variation seemed be to a more important factor in influencing the colour preference of jeans, rather than the widely reported chroma enhancement [3,16,43,60,61]. In this study, such an assumption was further validated. Fig. 2 illustrates the colour coordinates of the jeans in CAM16UCS under the experimental light sources (only four are plotted for clarity). It is obvious that when Duv values decrease from 0.02 to -0.02, the colours of the jeans get bluer (the coordinates in Fig. 2-c gradually shift towards the unique blue, i.e. a = 0, b < 0) and more saturated (the length of arrows gradually increase). Since the preference ratings generally increase with the decrease of Duv values, it is clear that, once again, people prefer bluer colours for jeans. As for chroma, in our previous work it was found that subjects generally preferred colours with lower chroma values (high CCTs) while in this case it was found that they generally preferred colours with larger chroma values (lower Duv values). So, definitely, the chroma variation might not be the influencing factor for the colour preference of the single-colour jeans. Meanwhile, it should be noted that the positive impact of the hue shift towards unique blue is only valid under certain conditions (i.e. for our latest work [30], from 2500 K to 5500 K; for this study, from Duv = 0.02 to Duv=5
Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Y. Liu, et al.
Fig. 2. Coordinates of jeans colours in CAM16UCS under different light sources: a) J-a plot, b) J-b plot, c) a–b plot.
0.01). When the CCT is too high (i.e. 6500 K) or the Duv value is too low (e.g. Duv < -0.01), the colours of the jeans appear to be too blue to be properly appreciated, as shown in Table 2.
3.3. Analysis of colour discrimination The data shown in Table 3 reveal that the observers’ colour discrimination performance for this study is obviously better than that of our former work [30]. Such a finding could be partially ascribed to the higher illuminance level adopted in the experiment. Moreover, in our past work light sources of different CCTs were adopted while in this study all the experimental lights were of 5500 K. Since our work has validated that 5500 K is the optimal CCT for colour discrimination [30], it is reasonable that the adjusted error scores of this work are generally smaller. From Table 3 we can conclude that for all whole colour regions, observers generally have a greater colour discrimination capability under the light with Duv value of -0.010 and suffer from poor colour discrimination under the light with Duv of 0.020. Such a statement is particularly valid for the hue region of blue jeans (mainly located in the Blue-Purple box). In addition, the correlation coefficient between the average value and standard deviation of the adjusted error scores is similar to those of our previous work [30] (r ≥ 0.80), indicating that when the observers exhibit better colour discrimination, their responses are more consistent and vice versa. Note that, for certain conditions, the optimal light source varies with different colour regions, thus the trends of adjusted error scores for different boxes under different light sources are not as consistent as those of the colour preference ratings. Moreover, it is quite clear in Table 3 that the adjusted error scores of the Green-Blue and Blue-purple boxes are remarkably larger than those of the other two boxes. Such a finding suggests that, for those colour regions, the colour discrimination attribute for lighting is of greater importance. In addition, the formerly mentioned rm-ANOVA approach was implemented again and it was found that in this condition the influence of the experimental lights, the colour regions (boxes), the gender of the subjects as well as the interaction between lights and colour regions were all significant (p < 0.005). When referring to the FM-100 Hue Test, the mechanism analysis regarding the causes of the error scores is of great importance. Some people believe that if a light source causes large colour differences between adjacent samples then the error scores will be low and vice versa. However, the results of our work [30] and those of Pardo et al. [8] denied that assumption by analyzing the correlation between the average colour differences and the final error scores. In this study, similar results were obtained, as shown in Table 4 Mean colour differences (CIEDE2000) between adjacent samples of the four FM-100 boxes. Red-Green
Green-Blue
Blue-Purple
Purple-Red
Overall
Duv
Min
Max
Avg
Min
Max
Avg
Min
Max
Avg
Min
Max
Avg
Avg
0.020 0.010 0.000 −0.010 −0.020 Overall
0.805 0.902 0.982 1.021 0.954 –
2.847 2.736 2.642 2.555 2.527 –
1.787 1.782 1.790 1.810 1.835 1.801
0.666 0.674 0.510 0.403 0.303 –
5.028 4.658 3.803 3.285 3.199 –
1.961 1.902 1.804 1.705 1.621 1.799
0.573 0.725 0.786 0.687 0.605 –
2.679 2.525 2.411 2.451 2.472 –
1.541 1.582 1.585 1.665 1.786 1.632
1.149 1.143 1.064 0.857 1.034 –
2.778 2.708 2.652 2.623 2.743 –
1.928 1.878 1.831 1.772 1.758 1.833
1.804 1.786 1.753 1.738 1.750 –
The maximum colour difference is underlined and bolded. Min: Minimum, Avg: Average, Max: Maximum. 6
Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Y. Liu, et al.
Fig. 3. Correlation between the hue difference of the FM-100 colour samples and the average adjusted error scores of the Hue test. Solid lines indicate the hue differences calculated in CAM16UCS (left axis), dot lines denote the average adjusted error scores of FM-100 test (right axis). For clarity purpose, only the data of two experimental lights and the jeans-relevant colours are shown while the results of other experimental sources are consistent.
Table 4. Although for the Red-Green box the mean colour differences and mean adjusted error scores correlated well (a larger average colour difference corresponds to a lower average adjusted error score), for the other three boxes no significant correlation is observed. Specifically, for the Blue-Purple box which contains most of the blue colours, the average colour differences increase with decrease of Duv values while the corresponding adjusted error scores decrease first from Duv = 0.020 to Duv=-0.010 and then increase. We believe that the failure of this average-colour-difference method in predicting error scores should be ascribed to its averaging operation, as well as to the scalar property of the colour difference measure. That is, the averaging of colour difference will lead to a great loss of individual colour difference information, while the scalar property of such a measure limits its performance in representing the visual differences presented with a colour gradient pattern, which is the core of the FM-100 test. Therefore, in this study a graphical analysis based on hue difference was conducted, as illustrated in Fig. 3. Despite of the comprehensiveness of the graphical manner, the main advantage of such an analysis method lies in the vector property of the hue difference measure, which is calculated by subtracting the hue value of a certain sample from that of its latter sample in CAM16UCS. For instance, as shown in Fig.3, the hue differences between #41 and #42 are negative values, indicating that there are transpositions in hue between these two samples under the experimental lights. Obviously, such information could not be demonstrated by colour difference. From Fig. 3, it is quite clear that the adjusted error score of the FM-100 test is closely associated with the hue difference of adjacent samples. In general, a larger hue difference corresponds to a smaller adjusted error score (e.g. #43-#45, #51-#59) and vice versa (e.g. #42 and #46-#48). The authors believe that this finding may provide an effective reference for future work, especially for the topic of establishing a quantitative measure to predict the colour discrimination capability of lighting.
3.4. Gender difference As stated earlier, the rm-ANOVA test has demonstrated that there is significant gender difference in colour perception, both for colour preference or colour discrimination. In this sub-section, detailed results are discussed. First, for the seven experimental jeans, the average colour preference ratings and standard deviations of male and female observers are illustrated in Fig. 4. It is quite clear that the rating trends of male and female observers are similar. However, for the lights with positive Duv values, the scores of males are significantly higher than those of females while for the negative-Duv conditions such
Fig. 4. Gender difference in the mean and standard deviation (SD) of preference ratings for the experimental jeans. 7
Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Y. Liu, et al.
Table 5 Gender difference in the intra-observer variability of our recent studies. Experiment
Amount of subjects
Abnormal rate (men)
Abnormal rate (women)
Jeans Exp 1 [30] Jeans Exp 2 Black and white [1] Fruit and Vegetables [1] Artificial Flowers [2] Oil Paintings [2]
30(M:15, 30(M:15, 30(M:16, 30(M:15, 36(M:17, 36(M:17,
1.9 % 2.8 % 6.3 % 0 17.6 % 0
7.6 % 9.5 % 7.1 % 6.7 % 15.8 % 10.5 %
F:15) F:15) F:14) F:15) F:19) F:19)
results are not found. In addition, from Fig. 4(b) it can be seen that, with positive Duv values, the standard deviations of male ratings are higher while for lighting conditions with negative Duv values, the opposite results are obtained. Such a finding indicates that, relatively speaking, male observers respond consistently when they prefer a certain light (e.g. negative Duvs) while female observers respond consistently when they dislike certain light (e.g. positive Duvs). Second, the gender difference in the intra-observer variability of this work was discussed, together with that of our former studies [1,2,30] with a balanced recruitment of male and female observers. As shown in Table 5, for most cases, the intra-observer variability of female subjects is significantly higher. This finding indicates that male observers have certain advantages in making stable colour preference judgements. Third, the gender difference in colour discrimination test is summarised in Fig. 5. From this figure, it is seen once again that female subjects exhibit better colour discrimination capability, which agrees with the results of our earlier study [30] and that of Huang et al. [62]. To sum up, we believe that our recent work on finding the optimal lighting for jeans has proven the existence of a gender difference in both colour preference and colour discrimination of lighting. In fact, for other related domains including genetics [63], neuroscience [64], ophthalmology [65] and biology [66], such a difference is also widely reported. Therefore, for future studies, we recommend to take this factor into consideration when designing an experiment or analysing appropriate data.
4. Conclusion Our recent work revealed that, under 200 lx, 5500 K was the best CCT for jeans lighting, which could provide optimal visual perception for colour preference and colour discrimination. In this study, based on similar experimental protocols, the optimal Duv value of 5500 K -500 lx lighting conditions was investigated. Our results recommend a Duv value of -0.010 for jeans lighting, by which the colour preference and colour discrimination simultaneously reach optimum once again. Based on colour appearance analysis, we have demonstrated that it is the hue shift that influences the colour preference of lighting for jeans. A graphical analysis approach was also proposed to better explain the causes of the error scores in the FM-100 Hue Test. In addition, significant gender difference was demonstrated from several new points of view. The authors believe that the findings of this study may contribute to a deeper understanding for display lighting, not only for jeans, but for other single-colour objects as well.
Fig. 5. Gender difference in the colour discrimination capability revealed by the mean adjusted error scores of the FM-100 Hue Test. 8
Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Y. Liu, et al.
Funding This work is supported by the National Natural Science Foundation of China (Project No. 61505149) and the Young Talent Project of Wuhan City of China (Project No 2016070204010111). References [1] Q. Liu, Z. Huang, M.R. Pointer, M.R. Luo, B. Wu, A. Liu, White lighting and colour preference, part A: correlation analysis and metrics validation based on four groups of psychophysical studies, Light. Res. Technol. (2019) in press. [2] Z. Huang, Q. Liu, S. Westland, M.R. Pointer, M.R. Luo, K. Xiao, Light dominates colour preference when correlated colour temperature differs, Light. Res. Technol. 50 (2018) 995–1012. [3] Y. Lin, M. Wei, K. Smet, A. Tsukitani, P. Bodrogi, T.Q. Khanh, Colour preference varies with lighting application, Light. Res. Technol. 49 (2017) 316–328. [4] T. Khanh, P. Bodrogi, Q. Vinh, D. Stojanovic, Colour preference, naturalness, vividness and colour quality metrics, Part 1: Experiments in a room, Light. Res. Technol. 49 (2017) 697–713. [5] M. Wei, K.W. Houser, What is the cause of apparent preference for sources with chromaticity below the blackbody locus? LEUKOS 12 (2016) 95–99. [6] G. Rui, Q. Wang, H. Yan, X. Shao, Investigation on factors to influence color emotion and color preference responses, Opt. – Int. J. Light Electron. Opt. 136 (2017) 71–78. [7] E. Mahler, J.J. Ezrati, F. Viénot, Testing LED lighting for colour discrimination and colour rendering, Color Res. Appl. 34 (2009) 8–17. [8] P.J. Pardo, M.I. Suero, Á.L. Pérez, G. Martínez-Borreguero, Optimization of the correlated color temperature of a light source for a better color discrimination, JOSA A 31 (2014) A121–A124. [9] P.J. Pardo, E.M. Cordero, M.I. Suero, Á.L. Pérez, Influence of the correlated color temperature of a light source on the color discrimination capacity of the observer, JOSA A 29 (2012) A209–A215. [10] L. Xu, M. Luo, M. Pointer, The Development of a Colour Discrimination Index, (2017). [11] W.A. Thornton, Color-discrimination index, JOSA 62 (1972) 191–194. [12] Z. Huang, Q. Liu, M.R. Luo, M.R. Pointer, B. Wu, A. Liu, The whiteness of lighting and colour preference, Part 2: A meta-analysis of psychophysical data, Light. Res. Technol. (2019) 1477153519837946. [13] Q. Liu, Z. Huang, M.R. Pointer, M.R. Luo, K. Xiao, S. Westland, Evaluating colour preference of lighting with an empty light booth, Light. Res. Technol. 50 (2018) 1249–1256. [14] M. Wei, K.W. Houser, Systematic changes in gamut size affect color preference, LEUKOS 13 (2017) 23–32. [15] Y. Tang, D. Lu, Y. Xun, Q. Liu, Y. Zhang, G. Cao, The influence of individual color preference on LED lighting preference, 49th Conference of the International Circle of Education Institutes for Graphic Arts Technology and Management (IC) and 8th China Academic Conference on Printing and Packaging, 2017, May 14, 2017 - May 16, 2017, Springer Verlag, Beijing, China, 2018, pp. 77–87. [16] M. Wei, K. Houser, A. David, M. Krames, Colour gamut size and shape influence colour preference, Light. Res. Technol. 49 (2017) 992–1014. [17] Q. Wang, H. Xu, F. Zhang, Z. Wang, Influence of color temperature on comfort and preference for LED indoor lighting, Opt. – Int. J. Light Electron. Opt. 129 (2017) 21–29. [18] P. Bodrogi, Y. Lin, X. Xiao, D. Stojanovic, T. Khanh, Intercultural observer preference for perceived illumination chromaticity for different coloured object scenes, Light. Res. Technol. (2015) 1477153515616435. [19] A. Liu, A. Tuzikas, A. Zukauskas, R. Vaicekauskas, Pi. Vitta, M. Shur, Cultural preferences to color quality of illumination of different artwork objects revealed by a color rendition engine, Photonics Journal, IEEE 5 (2013) 6801010-6801010. [20] T. Khanh, P. Bodrogi, X. Guo, P.Q. Anh, Towards a user preference model for interior lighting, Part 2: Experimental Results and Modelling, (2018). [21] T. Khanh, P. Bodrogi, Q. Vinh, D. Stojanovic, Colour preference, naturalness, vividness and colour quality metrics, Part 2: Experiments in a viewing booth and analysis of the combined dataset, Light. Res. Technol. 49 (2017) 714–726. [22] Q. Liu, Z. Huang, K. Xiao, M.R. Pointer, S. Westland, M.R. Luo, Gamut Volume Index: a color preference metric based on meta-analysis and optimized colour samples, Opt. Express 25 (2017) 16378–16391. [23] K. Smet, P. Hanselaer, Memory and preferred colours and the colour rendition of white light sources, Light. Res. Technol. 48 (2016) 393–411. [24] A. David, P.T. Fini, K.W. Houser, Y. Ohno, M.P. Royer, K.A. Smet, M. Wei, L. Whitehead, Development of the IES method for evaluating the color rendition of light sources, Opt. Express 23 (2015) 15888–15906. [25] K. Smet, W. Ryckaert, M.R. Pointer, G. Deconinck, P. Hanselaer, A memory colour quality metric for white light sources, Energy Build. 49 (2012) 216–225. [26] K. Smet, W.R. Ryckaert, M.R. Pointer, G. Deconinck, P. Hanselaer, Correlation between color quality metric predictions and visual appreciation of light sources, Opt. Express 19 (2011) 8151–8166. [27] K.G. Foote, M. Neitz, J. Neitz, Comparison of the Richmond HRR 4th edition and Farnsworth-Munsell 100 Hue Test for quantitative assessment of tritan color deficiencies, J. Opt. Soc. Am. a-Optics Image Sci. Vis. 31 (2014) A186–A188. [28] J. Seshadri, V. Lakshminarayanan, J. Christensen, Farnsworth and Kinnear method of plotting the Farnsworth Munsell 100-Hue test scores: A comparison, J. Mod. Opt. 53 (2006) 1643–1646. [29] T. Dan, H. Komatsubara, S. Kobayashi, N. Nasuno, Evaluation of Color Discrimination Under Led Lighting by Two Types of 100-Hue Test, Light-emitting Diode, (2013). [30] Q. Liu, Z. Huang, Y. Liu, M.R. Pointer, M.R. Luo, Q. Wang, B. Wu, Best lighting for jeans: Optimizing colour preference and colour discrimination with multiple correlated colour temperatures, Light. Res. Technol. (2019) in press. [31] F. Feltrin, F. Leccese, P. Hanselaer, K. Smet, Analysis of painted artworks’ color appearance under various lighting settings, 2017 IEEE International Conference on Environment and Electrical Engineering and 2017 IEEE Industrial and Commercial Power Systems Europe (EEEIC / I&CPS Europe) (2017) 1–6. [32] N. Narendran, Color rendering properties of LED light sources, Proceedings of SPIE - The International Society for Optical Engineering 4776 (2002), pp. 61–67. [33] T.Q. Khanh, P. Bodrogi, Q.T. Vinh, D. Stojanovic, Colour preference, naturalness, vividness and colour quality metrics, Part 2: Experiments in a viewing booth and analysis of the combined dataset, Light. Res. Technol. 49 (2016) 714–726. [34] S. Jost-Boissard, P. Avouac, M. Fontoynont, Assessing the colour quality of LED sources: Naturalness, attractiveness, colourfulness and colour difference, Light. Res. Technol. 47 (2014) 769–794. [35] S. Jost-Boissard, M. Fontoynont, J. Blanc-Gonnet, Perceived lighting quality of LED sources for the presentation of fruit and vegetables, J. Mod. Opt. 56 (2009) 1420–1432. [36] M. Islam, R. Dangol, M. Hyvärinen, P. Bhusal, M. Puolakka, L. Halonen, User preferences for LED lighting in terms of light spectrum, Light. Res. Technol. 45 (2013) 641–665. [37] T.Q. Khanh, P. Bodrogi, Colour preference, naturalness, vividness and colour quality metrics, Part 3: Experiments with makeup products and analysis of the complete warm white dataset, Light. Res. Technol. (2016) 1477153516669558. [38] P. Bodrogi, Y. Lin, X. Xiao, D. Stojanovic, T.Q. Khanh, Intercultural observer preference for perceived illumination chromaticity for different coloured object scenes, Light. Res. Technol. 49 (2015) 305–315. [39] T.Q. Khanh, P. Bodrogi, Q.T. Vinh, X. Guo, T.T. Anh, Colour preference, naturalness, vividness and colour quality metrics, Part 4: Experiments with still life arrangements at different correlated colour temperatures, Light. Res. Technol. (2017) 1477153517700705. [40] D.N.a.C.W. Jerome, Color rendering of light sources: CIE method of specification and its application, Illum. Eng. 60 (1965) 262–271. [41] J.P. Freyssinier, M. Rea, A two-metric proposal to specify the color-rendering properties of light sources for retail lighting, SPIE Optical Engineering +
9
Optik - International Journal for Light and Electron Optics xxx (xxxx) xxxx
Y. Liu, et al.
Applications, SPIE (2010) pp. 6. [42] M. Rea, NLPIP Lighting Answers : Light Sources and Color, Rensselaer Polytechnic Institute ; National Lighting Product Information Program, Troy, NY, 2004. [43] W. Davis, Y. Ohno, Color quality scale, Opt. Eng. 49 (2010) 033602-033602-033616. [44] K. Hashimoto, T. Yano, M. Shimizu, Y. Nayatani, New method for specifying color-rendering properties of light sources based on feeling of contrast, Color Res. Appl. 32 (2007) 361–371. [45] W.A. Thornton, Color-discrimination index, J. Opt. Soc. Am. 62 (1972) 191–194. [46] S. Fotios, G.J. Levermore, Perception of electric light sources of different colour properties, Int. J. Light. Res. Technol. 29 (1997) 161–171. [47] W.A. Thornton, A validation of the color-preference index, J. Illum. Eng. Soc. 4 (1974) 48–52. [48] M.R. Luo, The quality of light sources, Color. Technol. 127 (2011) 75–87. [49] K.A.G. Smet, J. Schanda, L. Whitehead, R.M. Luo, CRI2012: a proposal for updating the CIE colour rendering index, Light. Res. Technol. 45 (2013) 689–709. [50] K.A.G. Smet, W.R. Ryckaert, M.R. Pointer, G. Deconinck, P. Hanselaer, Memory colours and colour quality evaluation of conventional and solid-state lamps, Opt. Express 18 (2010) 26229–26244. [51] A. David, P.T. Fini, K.W. Houser, Y. Ohno, M.P. Royer, K.A.G. Smet, M. Wei, L. Whitehead, Development of the IES method for evaluating the color rendition of light sources, Opt. Express 23 (2015) 15888–15906. [52] T.Q. Khanh, P. Bodrogi, Q.T. Vinh, D. Stojanovic, Colour preference, naturalness, vividness and colour quality metrics, Part 1: Experiments in a room, Light. Res. Technol. 49 (2016) 697–713. [53] Q. Liu, Z. Huang, K. Xiao, M.R. Pointer, S. Westland, M.R. Luo, Gamut Volume Index: a color preference metric based on meta-analysis and optimized colour samples, Opt. Express 25 (2017) 16378–16391. [54] Q. Wang, H. Xu, J. Cai, Chromaticity of white sensation for LED lighting, Chinese Opt. Lett. 13 (2015) 073301. [55] A. Kevin, D. Geert, H. Peter, Chromaticity of unique white in object mode, Opt. Express 22 (2014) 25830–25841. [56] I. Acosta, Daylight Spectrum Index: Development of a New Metric to Determine the Color Rendering of Light Sources, (2017). [57] M.S. Rea, J.P. Freyssinier, White lighting: a provisional model for predicting perceived tint in “white” illumination, Color Res. Appl. 39 (2015) 466–479. [58] C. Li, Z. Li, Z. Wang, Y. Xu, M.R. Luo, G. Cui, M. Melgosa, M.H. Brill, M. Pointer, Comprehensive color solutions: CAM16, CAT16, and CAM16-UCS, Color Res. Appl. 42 (2017) 703–718. [59] R. Dangol, M.S. Islam, M. Hyvärinen, P. Bhusal, M. Puolakka, L. Halonen, User acceptance studies for LED office lighting: Preference, naturalness and colourfulness, Light. Res. Technol. 47 (2015) 36–53. [60] S. Jost-Boissard, P. Avouac, M. Fontoynont, Assessing the colour quality of LED sources: Naturalness, attractiveness, colourfulness and colour difference, Light. Res. Technol. 47 (2015) 769–794. [61] M. Wei, K.W. Houser, G.R. Allen, W.W. Beers, Color preference under LEDs with diminished yellow emission, LEUKOS 10 (2014) 119–131. [62] Y. Li, W. Shi, D. Li, F. Luo, X. Su, J. Yu, G. Cao, Study of healthy light-color parameters for LED lighting, Opt. – Int. J. Light Electron. Opt. 126 (2015) 4887–4889. [63] J.E. Vanston, L. Strother, Sex differences in the human visual system, J. Neurosci. Res. 95 (2017) 617–625. [64] S.E. Palmer, K.B. Schloss, J. Sammartino, Visual aesthetics and human preference, Annu. Rev. Psychol. 64 (2013) 77–107. [65] A. Panorgias, N.R.A. Parry, D.J. McKeefry, J.J. Kulikowski, I.J. Murray, Gender Differences in Peripheral Colour Vision; A Colour-Matching Study, Invest. Ophthalmol. Vis. Sci. 51 (2010) 2. [66] A.C. Hurlbert, Y. Ling, Biological components of sex differences in color preference, Curr. Biol. 17 (2007) R623–R625.
10