Open Access
Add to BibSonomy
Abstract
The correlated color temperature (CCT) of the monitor white needs to be controlled for the preferred image reproduction according to the surround lighting changes. The preferred display white prediction model according to the surround lighting color is proposed both for the emissive transparent display and opaque displays. To develop the model, the preferred CCT of the monitor white of a simulated emissive transparent display and an opaque display were investigated under four different surround lighting CCTs by conducting psychophysical experiments with twenty subjects.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Currently, displays are widely used under various lighting conditions. As the surround lighting conditions change, the display appearance also changes though the display color’s CIE chromaticities or the luminance level is the same. For example, Oh and Kwak [1] showed that 4,000 K light perceived as the bluish light when shown after being adapted to 3,500K while perceived as the yellowish light after being adapted to 5,000K white. Also it is known that the perceived display brightness is affected by the surround luminance level [2]. Previous findings indicate that although the display white appears bluish under a yellowish incandescent lamp, it looks yellowish under a bluish fluorescent lamp, and a monitor in a dark room looks brighter than when seen in bright sunlight. This change of appearance of the display occurs because of the chromatic and dynamic adaptation mechanism of the human visual system. Therefore, to maintain the same image quality regardless of the user environment, display colors must be adjusted according to the surround lighting condition changes. Especially with the widespread of mobile displays and the advent of the transparent display, whose image quality is strongly affected by the transmitted light, understanding the effect of the surround lighting on the display image quality becomes increasingly important.
All the existing data on the effect of the surround lighting colors on the display [3–6] follow the same rule: under higher correlated color temperature (CCT) lighting conditions, the display white must also be set to a high CCT, while a low display CCT must be set under a low CCT environment. For example, Choi and Suk [5] found that 6,500 K display white is preferred under around 3,000 K surround illuminance while 11,000K display white is preferred most under around 13,000 K surround illuminance.
If our eyes are fully adapted to the surround lighting, a display with the same CCT as the surround light will be perceived as neutral. However, a neutral-looking display CCT is not necessarily the same with the surround lighting CCT. This is because under surround lighting, the display watching conditions are mixed chromatic adaptation conditions, which means that our eyes are adapted partially to the surround light and partially to the display lighting. However, the exact relationship between the surround CCT and the preferred or neutral-looking display CCT is not known.
This study examines the effect of the changes of the display color appearance caused by the surround lighting color by investigating the preferred white of emissive transparent displays and opaque displays with different CCTs under various indoor surround lighting conditions. Based on the collected data, the preferred display white prediction model is proposed using the mixed chromatic-adaptation model between ‘prototypical display white’ i.e. ‘the preferred display white that the observer has in his/her mind’ and surround lighting.
2. Psychophysical experiment
Psychophysical experiments were conducted to collect data for the preferred white of the emissive transparent display and opaque display. The experiments were performed using 20 subjects and four different surround lighting conditions. Because emissive transparent monitors are not commercially available, a wide color gamut LCD monitor was used to simulate emissive transparent displays and opaque displays with various CCTs. For the emissive transparent display simulation, the scattering or diffraction of the transmitted image [7,8] is not considered since the experiment was done with the uniform background, which is not significantly affected by the other factors affecting the image quality of the transparent display.
Note that the term ‘emissive transparent display’ is used throughout this paper to distinguish our simulation from the transmissive transparent display. In the case of emissive transparent display, the resulting image on the display is the addition of the transmitted background scene with the self-luminous input image while in the case of transmissive transparent display, the background scene is used as the backlight of the input image.
2.1 Experimental setup
A 24-inch wide color gamut LCD display was used to simulate an emissive transparent display and an opaque display with various display CCTs. The LCD display had 2.2 monitor gamma and 240 cd/m2 peak white. The screen resolution was 1920 × 1200. Two floor lamps with LED bulbs whose luminance and chromaticity can be controlled were used to generate the various surround lighting conditions. The lamps were located behind and in front of the display. These locations were selected to increase the realism of the transparent display simulation by minimizing the shadow of the display and the reflection on the display. The walls behind the display were covered with matte white paper. Figure 1 shows the experimental setting shown at the participant’s position where the LCD monitor is simulating the 100% transparent display. The image simulation method is explained in Section 2.3.
2.2 Surround condition setting
The surround lighting conditions were set to have the similar luminance levels while CCTs are varying from 3,000 K to 6,000 K with a 1,000 K interval (four levels), which was the maximum achievable CCT range using the LED bulbs to have the high enough luminance levels for the average surround condition. Table 1 shows the actual measurement CIE colorimetric data of each surround lighting, which were obtained by averaging the measurement data measured at 15-points marked in Fig. 2(a). The surround luminance of the center position on the monitor was measured without the LCD monitor. Figure 2(b) displays an example of the surround luminance distribution. All the measurement for the surround and the monitor was conducted using the spectroradiometer CS-2000 with one-degree measurement area. Though each surround CCT has slightly different luminance levels as shown in Table1, the participants didn’t notice the luminance difference between the surround conditions.
Table 1. Average measurement CIE colorimetric data for each surround lighting condition
Fig. 2 Surround luminance measurement points (a) and example of surround luminance distribution (b).
2.3 Test display simulation
The simulated test displays—an opaque display and an emissive transparent display—were assumed to have an sRGB color gamut with 2.2 gamma. The luminance of the peak white of both test displays was fixed at 120 cd/m2 resulting in around 0.5 surround to display luminance ratio that belongs to the average surround condition and the chromaticity was simulated to vary from 3,000 K to 7,000 K with 500 K intervals. For the emissive transparent display, the transmittance was set to 40%, the transmittance level that the major display manufacturers such as Samsung Display and LG Display are demonstrating at the various display exhibitions. Figure 3 shows the chromaticities of the surround lightings and the simulated display white points in CIE u′v′ chromacity coordinates.
Fig. 3 Chromaticities of surround lightings and simulated display white points in CIE u′v′ chromacity coordinates.
Figure 4 shows the simulation process of the test displays. Note that all the simulations were conducted by manipulating RGB values of the original input image sets. Two types of input images were considered: test images and a background image. Five grayscale images and five color images were selected as test images, including outdoor scenes, colorful images, and human faces. All the images have the same native white point. The resolution of each image was the same as that of the LCD display. The background image was a photograph of the scene behind the display. To obtain the background scene, the camera was set at the same location as the participant’s eyes without the LCD monitor.
At first, RGB values of input background image (RGB_bg) were converted to CIE XYZ values (Step1) using GOG model [9] as display characterization model of the LCD monitor used in this experiment. The performance of the characterization model was evaluated by measuring 114 test colors and comparing the measured data with the predicted values. The average CIELAB color difference was 1.4 ΔE*ab. At Step2, the background image’s original CIE XYZ values calculated at Step1 were manipulated to simulate each surround lighting condition. Since the background image only contains the neutral colors, all the CIE XYZ values in the background image were simply transformed using the fixed ratio such as X' = αX, Y' = βY, Z' = γZ to reproduce the measured background chromaticity and luminance level.
Step3 represents the process to simulate the input images shown on the test displays. As explained before, all the test displays were assumed to have sRGB color gamut, 120 cd/m2 peak white and various white chromaticities. Using the given display color characteristics, input RGB values of the gray and color test images were converted to CIE XYZ images.
Step4 is the step to combine the background and input image. To simulate the emissive transparent display, CIE XYZ values of input image and 40% of CIE XYZ values of the background image were added together since the transmittance was assumed to be 40%. In the case of opaque display simulation, only input image CIE XYZ values were used. As the final step, combined CIE XYZ values were converted to RGB values to display on the LCD monitor using the inverse GOG model used for Step 1.
As a result, 90 display images (9 display CCTs × 10 test images) were generated for the opaque display and 360 (9 display CCTs × 4 surround CCTs × 10 test images) images were created for the emissive transparent display. Figure 6 shows examples of the simulated display images. Figures 5(a) and 5(c) show the 6,000K opaque displays under 3,000K and 6,000K surround conditions respectively and Figs. 5(b) and 5(d) are the 6,000K emissive transparent displays under 3,000K and 6,000K surround conditions.
Fig. 5 Examples of the simulated 6,000K display images: (a) opaque display under 3,000K surround, (b) transparent display under 3,000K surround, (c) opaque under 6,000K surround and (d) transparent display under 6000K surround.
2.4 Psychophysical experiment
Twenty university students (10 males and 10 females) in their early twenties with normal color vision participated in the psychophysical experiment. The experiment consisted of eight sessions (four surround CCTs × two display types). Each session’s experiment having the fixed surround CCT and display type was conducted independently and the experimental order of the sessions was randomly selected for each participant. Before starting each session’s experiment, the participant spent 5 minutes under a given surround condition for adaptation.
For the sessions using the emissive transparent display simulation, the participant was asked to adjust the position of the seat such that simulated background image was well aligned with the real background behind the display as shown in Figs. 5(b) and 5(d). The average distance was around 70 cm resulting in 41-degree viewing angle for the LCD monitor. The same distance from the LCD monitor and seat was used for the opaque display experiment. The participants were allowed to freely move their body during the experiment. During the emissive transparent display experiments, the participants reported that the display looked transparent though they knew that the image was the simulation.
The participant’s task was choosing the most preferred image among 9 rendered-images with the different display CCTs by freely increasing or decreasing the CCTs using a mouse wheel for each test image. CIE XYZ values of the selected display CCT were recorded for the data analysis. The order of the test images was randomly selected. Therefore, for each session, 10 most preferred display CCT points were selected from 10 different test images (five grayscale and five color images). Each participant made each judgment only once resulting in 80 selections.
3. Experimental result
Figure 6 summarizes the average of the selected CCTs of each test image under each surround condition, which indicates that the preferred display CCT is affected by surround condition, display type and also the image contents.
Since CCT is not perceptually uniform scale, for further data analysis, the recorded CIE XYZ values from all the participants were first averaged, and then the averaged XYZ values were used to calculate the corresponding display CCT.
Figure 7 compares the average preferred display CCTs for a test image displayed on the emissive transparent display and on the opaque display. The filled and empty circles represent color and grayscale images, respectively. The straight line represents the 45-degree line i.e. the same CCTs between two types of displays.
Fig. 7 Comparison between the preferred display CCTs for the transparent OLED display and opaque display under (a) 6,000 K surround, (b) 5,000K surround, (c) 4,000K surround and (d) 3,000K surround conditions.
First, it is noticeable that the preferred display CCTs decrease as the surround light CCT decreases regardless of the types of displays. In the case of 6,000K and 5,000K surrounds (Figs. 6(a) and 6(b)), most of the selected monitor CCTs are located between 5,000 and 6,000 K while the monitor CCTs between 4,000K and 5,000K are selected when the surround CCTs are 3,000K and 4,000K (Figs. 6(c) and 6(d)). Additionally, depending on the display type, the preferred display CCTs are affected differently by the surround CCTs. For 6,000 K surround lighting condition, as shown in Fig. 6(a), most of the data points are located below the 45-degree line meaning that a lower display CCT is preferred for the emissive transparent display than for the opaque display, while the reverse is observed for surround lighting under 3,000 K surround condition in Fig. 6(d).
The effects of the display type, image type, and the surround CCTs on the preferred display CCTs were further analyzed in detail considering the effect of the transmitted light on the emissive transparent display. Note that in the case of the emissive transparent display, the display CCT is the CCT of the self-luminous part of the display (self-luminous white CCT). However, participants were observing not only the self-luminous image but also the transmitted background image. Therefore, the total white CCTs, calculated using CIE XYZ values of the total white (the sum of the self-luminous white from the emissive transparent display with the average transmitted background light), were also used in the data analysis.
3.1 Difference of preferred display white CCTs according to display type and image type
Figure 8 compares the preferred CCT of the emissive transparent display with the preferred display CCT of the opaque display for the gray (left) and color (right) images. In the case of the gray images, the self-luminous white CCTs of the emissive transparent display are located above the 45-degree line while total white CCTs are lying on the 45-degree line. It is evident that not the preferred self-luminous white but the preferred total white CCT of the emissive transparent display is similar to the preferred display CCT of the opaque display. However, in the case of color images, not total white but the preferred self-luminous white CCTs of the emissive transparent display are similar to the preferred opaque display CCTs.
Fig. 8 Preferred self-luminous white CCT and total white CCT for the emissive transparent display compared to the preferred display CCT for the opaque display.
Since the difference of preferred display white CCTs between the gray and color images is not significant compared to the effect of surround lighting, both data is combined for further data analysis.
3.2 Changes of the preferred display CCTs according to surround lighting colors
The average preferred opaque display CCTs, self-luminous white CCTs and total white CCTs were calculated by averaging the data from all the test images for each surround lighting condition. Figure 9 summarizes the result. On average, similar CCTs were preferred for the opaque display and the total white (self-luminous + transmitted) of the emissive transparent display in spite of the luminance level differences. Note that the luminance of total white for the transparent display is higher than that of the opaque display. This result indicates that in the case of the emissive transparent display, CCT of the display must be controlled by considering not only self-luminous images, but also the transmitted background color information.
Regarding the effect of the surround CCT, a higher display CCT is preferred as the surround CCT increases, showing a linear relationship with lower slope than 1 (the 45-degree line). When the surround CCT is below approximately 5,600 K where the experimental data line and 45-degree line are crossing, the preferred display CCT is higher than that of the surround CCT, while the reverse is true for high surround CCT. Note that Choi and Suk’s studies [3,4] also showed a similar trend to that of Fig. 8, but with a different crossing point, i.e., the point that the monitor white and surround white are the same, near 10,000 K. Such a difference might originate from the different types of test images used in the experiment. Complex color images were used in this study while a full white screen was used to evaluate the preferred white in Choi and Suk’s studies, resulting in a higher preferred display CCT. Further investigation is needed to explain this discrepancy.
Figure 9 indicates that the observers had not fully adapted to the surround lighting conditions when they selected the preferred CCT image, although they had adapted to the surround light before watching the display images. If full adaptation had occurred, the preferred display CCT would have been similar to that of the surround light. The fact that the CCT range for the preferred display CCT is much smaller than that of the surround CCT implies that mixed chromatic adaptation occurred, which means that both the surround light and the display affected the chromatic adaptation of the observers.
4. Preferred display white prediction model according to the surround light color
4.1 Mixed chromatic adaptation model for preferred display CCT
It is well known that when people compare a soft-copy image on a display with a hard-copy image shown under illumination, the adaptation white point is shifted from the monitor white toward the paper white, resulting in the mixed adaptation condition [10–12].
According to Katoh et al. [10], mixed chromatic adaptation can be explained using (1), where LMSadaptation represents Ladaptation, Madaptation, or Sadaptation i.e. the resulting long, middle, or short cone signals of the adapting white point; LMSmonitor (Lmonitor, Mmonitor, or Smonitor) and LMSsurround (Lsurround, Msurround, or Ssurround) represent the cone signals for the monitor white and surround white, respectively. Radaptation is the ratio with which the monitor white contributes to the resulting adaptation point.
Note that (1) cannot be directly used to predict the preferred display CCT experimental result because the display white was not fixed in this study. The observers scanned the various display CCTs until they determined the preferred CCT using their own consistent judgment criteria. It seems that the observers are comparing the display white with the previously experienced preferred display white through their daily lives. Such a comparison mechanism is similar to the image quality judgment process of the familiar objects such as skin, green grass or blue sky (so called memory color) reproduced on the prints or displays for the preferred image reproduction [13].
Following the idea that “the basis of visual appreciation judgment is usually a comparison between the color sensations aroused by the reproduction, and a memory color i.e. mental recollection of the color sensations previously experienced when looking at objects similar to the ones being appraised” [14], Yendrikhovskij et al. [13] proposed a computational model of naturalness of a reproduced image. According to their model, the image quality judgment is made by evaluating the color difference between the ‘prototypical color’ – the mean value in appearance among objects of the corresponding object category – and the reproduced color.
In this study, we propose that the display white is also the color of the familiar objects since we are using the imaging devices under wide range of environments every day and we have our own judgment criteria for the preferred display white. Therefore, we can assume that during the preferred display white selection experiment the participants were choosing the display white that most resembles the ‘prototypical display white’ in their minds regardless of the surround conditions. However, the selected whites were different depending on the surround CCTs because of the different adaptation conditions.
Adopting the display white as memory color and the mixed chromatic adaptation model, it is assumed that mixed chromatic adaptation occurs between the chromaticity of the surround light and that of ‘prototypical display white’. Following this assumption, (1) can be rewritten as (2), where LMSmonitor is changed to LMSprototypical_white (cone signals of the prototypical display white).
Equation (2) can be applied to CAT02, i.e., the chromatic adaptation model used in CIE color appearance model, CIECAM02 [15], to calculate the corresponding CIE XYZ values of the selected monitor white under a reference viewing condition. If the observers were using consistent criteria to select the preferred display CCT, all the data points obtained under various surround conditions should result in the same corresponding colors under the reference viewing condition. Based on Fig. 9, it is assumed that ‘prototypical display white’ is 5,600 K, where the selected preferred monitor white is the same as the surround white.
4.2 Preferred display white prediction model
Based on (2), the preferred display white prediction model under a given surround lighting color is developed based on the CAT02 chromatic adaptation model used in CIECAM02 [15] as shown below:
Input: relative CIE XYZ values of the surround lighting and luminance of the adapting field, LA
- Surround lighting: Xsurround, Ysurround, Zsurround
- LA: Surround luminance (Luminance Y values in Table 1)
- Preferred Display White in reference condition: Xprototypical_white = 97.38, Yprototypical_white = 100.00, Zprototypical_white = 97.54 (5,600K based on this study)
Output: Predicted CIE XYZ values for the preferred monitor white under the given surround white: Xmonitor, Ymonitor, Zmonitor
Step1: Convert Xprototypical_white, Yprototypical_white, Zprototypical_white and Xsurround, Ysurround, Zsurround values to Lprototypical_white, Mprototypical_white, Sprototypical_white and Lsurround, Msurround, Ssurround cone signals using Eq. (3).
Step2: Calculate Adapting White Points Ladaptation, Madaptation, Sadaptation using (2) where Radaptation = 0.727 (Optimized using the experimental data)
Step3: Calculate the corresponding the preferred monitor white Lmonitor, Mmonitor, Smonitor signals under the given surround lighting condition
Step4: Calculate the corresponding the preferred monitor white XYZ signals under the given surround lighting condition using the inverse matrix in (3).
Step5 (only for the emissive transparent display): Calculate XYZ values of the transmitted light by multiplying the transmittance, T, of the display to XYZsurround then subtract the calculated XYZ from Step 4 result to obtain the self-luminous CCT for the emissive transparent display.
In this proposed model, it is assumed that only the chromaticities of the surround and the preferred display are affecting the degree of mixed chromatic adaptation. Therefore, in Step1, both Y values of surround lighting and the preferred display in the reference condition are set to 100. However, note that under different surround-display luminance ratio conditions, this assumption may not apply since this model is developed based on the experiments conducted under only one fixed condition.
The mixed adaptation ratio Radaptation in Step 2 was optimized for four surround conditions by minimizing the color difference between the experimental data and predicted display CCT. The degree of luminance adaptation D was set to 0.902, as calculated using the equation provided in CIECAM02 where F is set to 1 assuming the average surround condition. The calculation result showed that the optimized Radaptation value, calculated under four different surround lighting conditions, was 0.727. It is notable that the ratio of adaptation was also reported to be 0.7 by Kwon et al. [12], who compared the display images under two different surround lighting conditions.
Figure 10 compares the experimental data with the predicted preferred monitor white CCT using the newly developed model in this study, showing good agreement.
5. Conclusions
The effect of the surround light color on the preferred display white was investigated using a simulated emissive transparent display and an opaque display under four different surround CCT conditions. A psychophysical experiment was conducted using twenty subjects who were asked to select the most preferred image among images rendered to have various display CCTs for each test image.
The result showed that the preferred display CCT of the opaque display was similar to that of the emissive transparent display including the transmitted light, implying that the image quality was determined by the final output color shown on the display regardless of the display type. A linear relationship was observed between the selected preferred display CCT and the surround light CCT. A higher surround light CCT corresponded to a higher preferred display CCT confirming the findings with the previous studies. When the surround light has 3,000 K, a display of approximately 4,700 K was preferred, while under 6,000 K light, a display of approximately 5,700 K was preferred.
The change of the preferred display CCT according to the surround light chromaticity is explained by using the mixed chromatic adaptation model between the surround light and ‘prototypical display white’, assuming that each subject used the ‘prototypical display white’ as the reference display white for the chromatic adaptation. The proposed model performs well when ‘prototypical display white’ is set to 5,600 K and the adaptation ratio with which the monitor white contributes to the resulting adaptation point, R, is set to 0.727.
In spite of the good performance of the proposed model, it should be noted that the experiment was conducted in the limited setting in this study. Since the display color appearance is affected not only by CCTs of the monitor and surround but also many other factors such as the display luminance level and surround-monitor luminance ratio, further extensive experiments are required on this topic to evaluate the proposed mixed chromatic adaptation model based on prototypical display white.
Funding
Samsung Display Co., Ltd.
References
1. S. Oh and Y. Kwak, “Hue and warm-cool feeling as the visual resemblance criteria for iso-CCT judgment,” Color Res. Appl. (2018), doi:. [CrossRef]
2. Y. S. Baek, Y. Kwak, and S. O. Park, “Monitor brightness changes under a wide range of surround conditions,” J. Opt. Soc. Am. A 34(2), 216–223 (2017). [CrossRef] [PubMed]
3. Y. J. Cho, “Device and method for auto-adjustment of image condition in display using data representing both brightness or contrast and color temperature,” U.S. Patent No 6,292,228 (2001).
4. S. H. Lee, J. H. Koo, D. K. Choi, K. I. Song, K. R. Kwon, S. K. Jeon, and B. G. Kim, “Surrounding light judging method and video compensation control apparatus using the same,” U.S. Patent No 6,822,695 (2004).
5. K. Choi and H. J. Suk, “User-preferred color temperature adjustment for smartphone display under varying illuminants,” Opt. Eng. 53(6), 061708 (2014). [CrossRef]
6. K. Choi and H. J. Suk, “Assessment of white for displays under dark- and chromatic-adapted conditions,” Opt. Express 24(25), 28945–28957 (2016). [CrossRef] [PubMed]
7. H. J. Kwon, C. M. Yang, M. C. Kim, C. W. Kim, J. Y. Ahn, and P. R. Kim, “Simulation of blur in transmitted image through transparent plastic for transparent OLEDs,” J. Disp. Technol. 12(8), 851–858 (2016). [CrossRef]
8. Z. Qin, J. Xie, F. C. Lin, Y. P. Huang, and H. P. D. Shieh, “Evaluation of a transparent display’s pixel structure regarding subjective quality of diffracted see-through images,” IEEE Photonics J. 9(4), 1–14 (2017). [CrossRef]
9. R. S. Berns, “Methods for characterizing CRT displays,” Displays 16(4), 173–182 (1996). [CrossRef]
10. N. Katoh, K. Nakabayashi, M. Ito, and S. Ohno, “Effect of ambient light on the color appearance of softcopy images: Mixed chromatic adaptation for self-luminous displays,” J. Electron. Imaging 7(4), 794–806 (1998). [CrossRef]
11. S. A. Henley and M. D. Fairchild, “Quantifying mixed adaptation in cross-media color reproduction,” in Proceedings of Color and Imaging conference, (Society for Imaging Science and Technology, 2000) 305–310.
12. O. S. Kwon, T. Y. Park, and Y. H. Ha, “High fidelity color reproduction of plasma displays under ambient lighting,” IEEE Trans. Consum. Electron. 55(3), 1015–1020 (2009). [CrossRef]
13. S. N. Yendrikhovskij, F. J. J. Blommaert, and H. De Ridder, “Color reproduction and the naturalness constraint,” Color Res. Appl. 24(1), 52–67 (1999). [CrossRef]
14. R. W. G. Hunt, Reproduction of Colour 5th Ed. (Fountain, 1995), Chap. 5.
15. CIE, CIE 159:2004 A Colour Appearance Model for Colour Management Systems: CIECAM02 (CIE, 2004).
