Black perception (perceived blackness of gray 0) of transparent OLED displays was studied in this paper. In pre-test, maximum luminances of acceptable black level under various surround conditions were found in a non-transparent display. In the first experiment, the luminance of a transparent patch was compared with that of an opaque one in order to find the effect of transparency on black perception. As a result, participants perceived the transparent patch darker than the opaque one even when the two were in similar luminance levels, which we termed as the “Transparency Effect.” In the second experiment, the perceived brightness of gray 0 with various background brightness conditions was investigated to observe the effect of induced black perception. Most participants perceived the luminance of gray 0 darker with brighter background luminance, but some did not. It might result from transparency of gray 0 which had a role as a window presenting the area overlapped with a transparent OLED display.
© 2017 Optical Society of America
Black level is an important factor in perceived image quality of digital displays. When manufacturers promote their devices, they use contrast ratio (the ratio between the luminance of the black and white) as one of the key indices of display quality. To some extent, it is true; In a previous research , participants presented with same two pictures with different black levels, picked the one with deeper black as the higher quality one. Also, Mantiuk and colleagues  explored the luminance of ‘absolute black’, the level of darkness which is always perceived as black under any surround luminance.
Perceived blackness can be directly influenced by surrounding luminance. Previous researches [2–5] revealed that the brighter the surround, the darker the perceived blackness. Participants felt deeper black when the contrast ratio of black stimulus to surround luminance was more than 1:100 . Although these papers [3–5] called the area encircling black test stimulus as the “surround”, the concept is similar to the “background” in Color Appearance Model (CAM). The “surround” conditions in CAM means exterior lighting conditions over visual angle of 10° of display area. Therefore, this paper will call the area encircling test stimulus “background” and the peripheral area of exterior lightings as “surround”.
The “surround” luminance (ambient luminance) also affects the perceived blackness. Park and Fairchild  mentions that dark gray, not black, could be perceived as black under a certain surround luminance. In the previous paper, the brighter the surround, the participants perceived deeper blackness . In strict sense, however, this paper did not show the direct relation between the surround luminance and the perceived blackness in that they simulated the ambient condition by adjusting the display luminance around a black target. Although the field of surround in this paper reached to about visual angle of 20°, it hardly seemed to simulate an ambient environment. Controlling the brightness of real exterior lightings is required in order to observe the direct relation between the surround luminance and the perceived blackness.
In order for consistent calculation results in CAM, it is crucial to find the “black-point”, which is the gray range perceived as black under various surround conditions. In the case of transparent displays, however, black perception is more difficult to define. Electronic devices assign discrete levels of brightness of images (gray levels) in digital values from 0 to 255, in which the black is set to level 0. This principle is applied to transparent displays as well. However, it should be noted that the perceived blackness of gray 0 in transparent OLED displays needs to be understood differently than that in transparent LCD displays. Although the latter does use ambient light as backlight, but viewers are able to perceive black as pixels actually block backlight at gray 0 (LCD pixels are light valves). When surround luminance increases, the luminance of gray 0 stays the same while that of gray 255 becomes higher. This, in turn, improves the contrast ratio, which makes gray 0 look even darker. On the other hand, transparent OLED displays have self-luminous pixels and do not have control over transmitted light; The amount of transmitted light is always the same regardless of the screen content. Therefore, the total amount of luminance is the sum of the light from pixels and the transmitted light. If the amount of transmitted light increases, the luminance value of “gray 0” also rises, and this makes more difficult for viewers to perceive gray 0 as black. This is why black perception in transparent OLED displays should be studied differently than other normal opaque displays.
This paper focuses on the black perception (perceived blackness of gray 0) in transparent OLED displays. For better comparison, the maximum luminance accepted as “black” was found in a normal (non-transparent) display under various surround conditions as a pre-test; Our initial goal was to find the approximate range of luminance of “black” as perceived by viewers. Next, we measured perceived brightness of gray 0 in a transparent OLED display. If “transparency” affects the perceived brightness, it may no longer be the same as the sum of luminance from pixels and the transmitted light. If transparency is not a factor, the brightness of gray 0 must be the same as the transmitted surround luminance in a transparent OLED display. Finally, we explored the effect of background luminance on perceived brightness of gray 0. Generally, gray 0 with brighter background is perceived darker in a normal display. We explored whether this was the same with transparent OLED displays as well.
2. Pretest: acceptable luminance as black in a normal display
Experimental design was a within-subject design. Independent variables were surround luminance and congruence between white-points of surround lightings and a LCD monitor. Congruent condition had same white-point between surround and monitor, but non-congruent condition did not. The white-points corresponded to three common color spaces: D50, D65, and DCI-P3. The values of the white-point in CIE1931 were presented in Table 1. There were 8 levels of surround luminance: 10, 20, 30, 40, 50, 100, 150, and 200% of maximum luminance of a LCD monitor (230cd/m2). The dependent variable was maximum luminance of a test patch being perceived as black. Luminance of the test patch had eleven levels: 1~10% (each 1% step) and 12% of maximum luminance of the monitor.
The given task for participants was to select the brightest patch of acceptable black (method of adjustment). A patch was presented in the full size of a 27-inch monitor (model name: HP DreamColor Z27x) (Fig. 1). The viewing distance was 60cm and the size of the stimulus covered visual angle of 60°.
We performed a repeated-measured Analysis of Variance (ANOVA) on the data from two participants (including authors) using Minitab 16. The main effect of surround luminance was significant (F(7, 127) = 109.15, p < 0.001), but the effect of congruence was not (F(1, 127) = 0.19, p = 0.667) (Fig. 2). The interaction effect between surround luminance and congruence was not significant (F(7, 127) = 0.12, p = 0.997). In addition, analyzing the effect of white-point of a Monitor and surround lightings, the main effects of white-point were significant (Monitor: F(2, 71) = 3.57, p < 0.05; Surround: F(2, 71) = 3.35, p < 0.05). The effect size of surround luminance was bigger than those of white-points. The interaction effect between surround luminance and white-point of surround lightings was also significant (F(14, 71) = 2.01, p < 0.05), but the interaction between surround luminance and white-point of a monitor was not (F(4, 71) = 0.43, p = 0.822).
As a result, maximum luminance of a patch being acceptable as black, depended on the brightness of surround luminance, which is a similar result to those from previous researches [2,6]. However, luminance of acceptable black decreased under the brightest condition (a black arrow in Fig. 2). It may be due to the reflected light of surround luminance. The monitor screen reflects surround luminance of exterior lightings according to reflectance of a screen. The monitor used in the pre-test had 1.4% of reflectance measured directly in front. When surround luminance increased, the amount of reflected light also increased which affects black perception [7,8]. While the contrast ratio of a black stimulus to surround luminance increases, the luminance of acceptable black does not increase due to reflectance of a monitor.
In this test, average maximum luminance of acceptable black was 12cd/m2 under a bright surround condition (345cd/m2). Looking at each individual data, the maximum luminance of acceptable black was less than 18cd/ m2. This value will be used as a reference point when comparing the acceptable black between a normal (opaque) and a transparent display.
3. Black (gray 0) perception in a transparent OLED display
Applying the result of the pre-test to transparent OLED display, it seems to be more difficult for the participants to perceive black in a transparent OLED display. This is because the luminance of gray 0 in a transparent OLED display can be higher than 18cd/m2 with high transparency and surround luminance level. However, we do not know the effect of transparency on perceived brightness. As gray0 becomes a window showing the area overlapped with a display, transparency has a definite influence in black perception, which we investigated by comparing the test patch (transparent) with the reference patch (non-transparent) in experiment 1 under various surround conditions.
In order to reflect a real environment in the experiment, we measured the luminance of various spots in a road shop and in an arcade near a subway station (Fig. 3). We selected these two places because a transparent OLED display is likely to be used as a public information display (PID) for advertising . The measured brightness was from 70 to 360cd/m2. We did not choose luminance below 100cd/m2 since there is no merit in using transparent display in dark areas where there is little light transmitting through the panel. Finally, we selected three points 150, 250, and 350 cd/m2 for surround conditions of the experiment 1.
Experimental design was a within-subject design. Independent variables were correlated color temperature (CCT) of surround lightings and surround luminance. Levels of CCT were three: 5500K, 6500K, and 13000K. Levels of surround luminance were three: 150, 250, and 350 cd/m2. All conditions of surround luminance were average surround conditions in Color Appearance Modeling, which were included within the range of 20~99% of maximum luminance of the display.
The task given to the participants was to adjust the luminance of a test patch (transparent) to match the brightness of a reference patch (non-transparent). The luminance of a reference patch corresponded to that of gray0 under various surround conditions, which is the level of transmitted light (Table 2). To eliminate bias, the starting luminance of a test patch was given from two points: below and above the luminance of the reference patch. The size of the test and the reference patch was visual angle of 2°. To make the non-transparent reference patch, a small non-transparent film was attached to the back of the display. Each patch had imagery transparent background with visual angle of 10° (Fig. 4). This experiment again was given from the brightest to the darkest condition. The transparent display unit used in this experiment was a prototype built by Samsung display for R&D purpose. It had transmittance of about 30%.
Thirteen subjects (9: female, 4: male) participated in the 13000K condition, twelve (9: female, 3: male) in the 5500K condition, and eight (5: female, 3: male) in the 6500K condition. As a result, the participants adjusted the luminance of a transparent test patch given brighter than that of a non-transparent reference patch. To validate the difference between the luminance of gray0 and those of test patches, we performed the one-sample T-test using minitab 16. The standard value was the luminance of each gray 0 under various surround conditions. All conditions were statistically significant (p < 0.001). An average luminance of a test patch was presented under each surround luminance (Fig. 5 and Table 3). Participants did not judge the brightness of a test and a reference patch on the bases of absolute luminance. If absolute luminance was the bases of judgment, they should select a test patch of gray 0 because absolute luminance of gray 0 was similar to a reference patch in all conditions. However, they chose the patch with self-luminous light which was brighter than gray 0. We called this the “Transparency Effect.” This phenomenon appeared in all CCT of surround lightings conditions although the effect of CCT of surround lightings was marginally significant (F(2, 188) = 3.04, p = 0.05). It seemed that they perceived a test patch of gray0 as a window or being nothing, not as a stimulus. A test patch was perceived disappearing when gray0.
A similar effect was shown in space brightness perception. People tend to perceive the transmitted light through a window darker. When participants judged the space brightness between the space with transmitted light through window and the space without a window, they perceived the latter brighter than the former in the Yamada and colleagues’ research . This effect was enhanced by a scenic view. It seemed that a scenic view strengthened the feeling of perceived transmitted light through a window. Applied in a transparent OLED display, people will perceive a transparent patch darker than a non-transparent patch even when the two patches have the same luminance. As a result, participants perceived the same level of luminance from a non-transparent patch and a brighter transparent patch. We concluded that the transparency affected the brightness perception of a patch. The exact cause of this effect is not yet known. We assume that people develop pre-conceived bias on the brightness of transmitted light when they notice the existence of a window. Underlying biological mechanisms will be studied further.
3.3 Additional experiment: with a color background
We observed Transparency Effect under the condition where there are objects behind a transparent display. In order to simulate this condition, we attached a color patch on the back wall of a lighting booth. A color patch was positioned within the background of a transparent test patch (Fig. 6). Each participant saw different colors through a test patch. The position of the eyes was slightly different because a sitting height was different although the height of a chair and the viewing distance were same. Surround luminance was 150cd/m2 and CCT was 5500K. Luminance of gray0 was 35.02cd/m2, and the reference patch (non-transparent) was set to the closest value 31.92cd/m2. The procedure was the same as in experiment 1.
Eight subjects (5: female, 3: male) were participated in the additional experiment including the authors. The Transparency Effect was also observed. An average luminance of a test patch was 85.56cd/m2 ( ± 29.29), which was the result of matching the brightness with the reference patch. We performed one-sample T-test using minitab 16. The average luminance of a test patch was statistically different from that of the reference patch (t(23) = 2.67, p < 0.05). Luminance of a test patch did not exactly reflect the real brightness of a test patch. We measured luminance of a test patch directly in front, but the eye position of each participant was different, which might have resulted in higher luminance values than those from experiment 1. Namely, the degree of the Transparency Effect is likely to vary according to the color of an object behind a transparent display.
4. Induced blackness by background luminance in a transparent OLED display
In a transparent OLED display, it is hard to have acceptable black with same physical intensity of light to a non-transparent display. Luminance of gray0 in a transparent display increases with surround luminance while luminance of acceptable black did not increase by more than 18 cd/ m2 in a non-transparent display. In addition, transparent effect makes black (gray0) being perceived as a window. In order to explore feasibility of enhancement of black perception in a transparent display, we measured the induced blackness by background luminance. In this experiment, background was defined as the area encircling a gray0 stimulus.
4.1 Pre-test: in a normal display
Before the experiment in a transparent display, we observed induced blackness in a normal display. Induced blackness is a well-known effect, but our purposes were to obtain the decrease trend quantitatively and to determine Experimental the procedure.
Surround luminance was 350cd/m2. CCT of a monitor and surround lightings was the same at 5500K. Luminance of a target patch (center area) was 69.34cd/m2 of gray145 which was not set to gray0, because the brightness of gray0 in a transparent display is not zero. We set 6 levels of background luminance at between 100%-200% of the luminance of a center area (115%, 130%, 145%, 163%, 180%, and 200%). Due to limitation of luminance of a monitor, the condition above 200% was not included.
We presented a certain level of luminance as a reference (instead of 69.34) on the black background. The reference had the numeric value, 60. The absolute value was not important because we calculated the relative change. Then, a target patch with background was presented. A task was to rate the brightness of a target patch (the numeric value) relative to the reference. The reference patch was presented only once before presenting test patches.
The size of the target was visual angle of 2°. A background was presented in the full size of a 27-inch monitor which was used in the pre-test of experiment 1. The viewing distance was 60cm and the size of the stimulus covered visual angle of 60°. The order of background luminance was randomized. Two participants (including one author) conducted three blocks, a total of 12 trials (Fig. 7). A first session was a practice, so their data was not used to analyze.
We performed the simple regression using minitab 16. The predictor was the background luminance (cd/m2). The decreasing trend in the brightness of the target patch appeared, but the result of regression was not statistically significant (p = 0.085) (Fig. 8). Participants felt difficult rating a test patch comparing the reference patch. They tended to rate it based on brightness of the previous test patch. As the result of the pre-test, we decided not to use numeric rating system of brightness. Instead we asked participants to adjust luminance of a patch in the main experiment. Furthermore, we set the range of background luminance to above 200% of a test patch.
Experimental design was a within-subject design. Independent variables were surround luminance and background luminance. We set 4 levels of surround luminance: 150, 250, and 300cd/m2. All conditions of surround luminance were average surround conditions in Color Appearance Modeling, which were included within the range of 20%-99% of maximum luminance of the display. The brightest surround condition was changed from 350cd/m2 to 300cd/m2. When 350cd/m2, a light booth was not stable after the experiment 1. To maintain a stable condition during the experiment 2, we reduced the luminance of the brightest condition.
CCT of surround was 13000K. There were 15 levels of surround luminance: 110%-250% of luminance of gray0 under each surround with a step size of 10%. The task was to adjust the luminance of test patch (non-transparent) to match the brightness with the reference patch of gray0 (transparent). Because the luminance of a transparent patch did not decrease below luminance of gray0, a separate test patch was made as a non-transparent one. The luminance of a reference patch was the same with that of gray0. To eliminate bias, the starting luminance of the test patch was randomized: one was under luminance of a reference patch and the other above. The size of the test and the reference patch covered visual angle of 2° (Fig. 9). To make the non-transparent reference patch, a small non-transparent film was attached on the back of the display. Two patches had transparent backgrounds with visual angle of 10°, but the background of the reference patch (transparent) had self-luminous light. The experiment started from a brightest condition and ended with the darkest.
We performed a repeated-measured Analysis of Variance (ANOVA) on the data of sixteen participants (12: female, 4: male) using Minitab 16. The main effect of background luminance was significant under each surround luminance (150cd/m2: F(14, 448) = 4.28, p < 0.001; 250cd/m2: F(14, 150) = 8.11, p < 0.001; 300cd/m2: F(14, 447) = 12.61, p < 0.001). The brighter was background luminance, the darker did participants perceive gray0 (Fig. 10). For example, luminance of gray0 under 150cd/m2 of 13000K surround luminance in the experiment 1 was 37.83cd/m2 and perceived brightness of gray0 with the same condition was 61.69cd/m2. Perceived brightness of gray0 with 250% background luminance decreased to 18.1cd/m2. This trend, induced blackness, was same to a non-transparent display . However, minimum value of brightness of induced blackness was brighter than maximum luminance of acceptable black in a non-transparent display.
Looking at each individual data, however, some participants did not have this trend. Namely, they did not perceive a gray0 patch as the darker black although background luminance increased. We divided participants to two groups: a change group and a flat group. Among sixteen participants, five participants were included in a flat group. They adjusted luminance of a test patch as a same level regardless of increase of background luminance. Excluding data of the flat group, the trend of induced blackness in the change group appeared clearly.
To explain the perceived brightness of gray0 with the change in background luminance, a linear regression was performed separating date between two groups. The predictor was background luminance. The value of the background luminance was converted from percentage to luminance (cd/m2). The result of regression in the change group was shown in the Table 4. In the change group, the perceived luminance of a gray0 decreased with increase in background luminance.
However, the trend seemed to converge if the background luminance was more than 180% of gray0 (Fig. 11). To validate this change-point, a two-line regression model  was performed. It was based on an F-test, which was using residual sums of squares (RSSs). The section of the first slope was 110%-170% and one of the second slope 180%-250%. As a result, the two-line regression was significant in each surround luminance (150cd/m2: F(1,327) = 4.87, p < 0.05; 250cd/m2: F(1, 327) = 2.53, p < 0.001; 300cd/m2: F(1, 327) = 21.88, p < 0.001).
The result of regression in the flat group was not significant (150 cd/m2: F(1,148) = 0.20, p = 0.656; 250 cd/m2: F(1, 148) = 0.08, p = 0.772; 300 cd/m2: F(1, 148) = 0.33, p = 0.567). They perceived similar bright in the patch of gray0 regardless of background luminance (Fig. 12). It may be because that gray0 had a role as a window presenting the area overlapped with a display. In a non-transparent display, people do not separate a center area from background when they rate the brightness of the center area. Simultaneous contrast is the classic example of this effect. People perceive the center area with a brighter background darker than that with darker background. However, in a transparent display, some people separated a center area from background, which was a window. It was a unique effect occurring only with transparent OLED display.
We explored black perception (perceived blackness of gray 0) in transparent OLED display in comparison to non-transparent display (or area). Participants had difficulty perceiving black in a transparent OLED display because gray0 has a role not only as black, but also as a window. Similar results were reported [9, 13, 14] where people did not perceive gray0 of test images as sufficient black. Kwak and colleagues  calculated brightness in a simulated transparent OLED. With transmittance of 10%, the lightness tone curves were almost the same in the preferred gamma between transparent and non-transparent displays . In the subsequent experiments , they asked participants to select a favorite image among various gamma values under varying transmittance and surround conditions. As a result, participants chose 1.7 monitor gamma as the most preferred value when simulating a transparent OLED display with 50% of transmittance. Comparing the CIELAB L* distributions of test images on a non-transparent display with 2.0 monitor gamma with those of a transparent display with 1.7 monitor gamma, the value of L* of black (gray0) was different between 0 and 35 although two images had similar lightness tone curves. They just perceived gray0 of preferred monitor gamma relatively darker comparing those of different monitor gammas. They might have sensed gray0 not as black, but as dark gray. Kwak et al proposed the concept, “SR,” which was the relative ratio of black (gray0) to highest luminance of a monitor. If transmittance increase or surround luminance becomes brighter, SR increases and people hardly perceived gray0 as black. This implies that representing sufficient black is important for image quality, but it is not easy accomplish a transparent OLED display.
Hwang and colleagues  predicted that reproduction of black was difficult only with a transparent display, and insisted that an additional structure was needed to block the exterior lights. When there is no transmitted light, however, people cannot see screen contents and the area behind a transparent display simultaneously. Therefore, further study is needed to suggest the method that enhances black perception in a transparent display.
We note several limiting factors regarding this study. First, a larger pool of participants and display units may be needed for better statistical generalization. Second, the main purpose of this paper is to investigate the existence of Transparent effect. Further study will be needed to actually quantify its affects on black perception and induced blackness. Third, in order to obtain more accurate data related to whether the participants are looking at the overlapped region between the display and the patch, tracking their eye movements as well as focal points in the depth direction will be helpful. Nevertheless, the significance of this paper may be meaningful in that the first try for black perception in a transparent OLED display measuring physical values of luminance.
References and links
1. T. Eda, Y. Koike, S. Matsushima, T. Ishikawa, K. Ozaki, and M. Ayama, “Influence of blackness on visual impression of color images,” Kansei Eng. Int. J. 10(1), 49–58 (2010). [CrossRef]
2. R. Mantiuk, S. Daly, and L. Kerofsky, “The luminance of pure black: exploring the effect of surround in the context of electronic displays,” in Proceedings of Electronic Imaging 2010 (EI, 2010), pp. 75270W.
5. T. Eda, Y. Koike, S. Matsushima, K. Ozaki, and M. Ayama, “Influence of surround luminance upon perceived blackness,” in Proceedings of Electronic Imaging 2009 (EI, 2010), pp. 72400T.
6. S. J. Park and M. Fairchild, “Color reproduction using “Black-Point Adaptation,”” in Proceedings of Color and Imaging Conference (CIC, 2001), pp. 245–250.
8. K. Chakrabarti, R. V. Kaczmarek, J. A. Thomas, and A. Romanyukha, “Effect of room illuminance on monitor black level luminance and monitor calibration,” J. Digit. Imaging 16(4), 350–355 (2003). [CrossRef] [PubMed]
9. C. S. Hwang, S. H. Park, and K. I. Cho, “Technical trends and prospects in transparent display,” Electron. Telecommun. Trends 25(5), 20–32 (2010).
10. S. Yamada, R. Tanaka, H. Shinoda, and Y. Seya, “Space Brightness Affected by a Scenic View through a Window,” In Proceedings of Association Internationale de la Couleur 2015 (AIC, 2015), pp. 1211–1216.
11. K. Shinomori, Y. Nakano, and K. Uchikawa, “Influence of the illuminance and spectral composition of surround fields on spatially induced blackness,” J. Opt. Soc. Am. A 11(9), 2383–2388 (1994). [CrossRef] [PubMed]
12. S. A. Julious, “Inference and estimation in a changepoint regression problem,” J. Royal Statistical Soc.: Series D 50(1), 51–61 (2001).
13. Y. Kwak, Y. S. Baek, and J. C. Kim, “Optimal tone curve characteristics of transparent display for preferred image reproduction,” in SID Symposium Digest (2014), pp. 1123–1126.
14. Y. Kwak, H. Ha, S. Lee, H. Kim, Y. J. Seo, and B. Yang, “Optimal monitor gamma of transparent display,” in SID Symposium Digest of Technical Papers (SID, 2015), pp. 196–199.