Open Access
Add to BibSonomy
Abstract
We selected appropriate indices for color rendition and determined their recommended values for ultra-high-definition television (UHDTV) production using white LED lighting. Since the spectral sensitivities of UHDTV cameras can be designed to approximate the ideal spectral sensitivities of UHDTV colorimetry, they have more accurate color reproduction than HDTV cameras, and thus the color-rendering properties of the lighting are critical. Comparing images taken under white LEDs with conventional color rendering indices (Ra, R9-14) and recently proposed methods for evaluating color rendition of CQS, TM-30, Qa, and SSI, we found the combination of Ra and R9 appropriate. For white LED lighting, Ra ≥ 90 and R9 ≥ 80 are recommended for UHDTV production.
© 2017 Optical Society of America
1. Introduction
Ultra-high-definition television (UHDTV) is the next-generation television standard specified in Recommendation ITU-R BT.2020 (Rec.2020) [1]. Its resolution is twice and four times higher than high-definition television (HDTV) specified in Recommendation ITU-R BT.709 (Rec.709) [2] and its frame frequency is specified up to 120 Hz. The system colorimetry of UHDTV has wider color gamut than that of HDTV because its three primary color points are set at wavelengths of 630 nm, 532 nm, and 467 nm on the spectral locus. Because ideal spectral sensitivities calculated by linear transformation of color matching functions based on Rec. 2020 primary colors have less negative lobes than those of Rec. 709, UHDTV cameras are more suitable for simulation of ideal spectral sensitivities using dichroic prisms. Thus, the spectral sensitivities of our UHDTV camera were designed to be very close to the ideal spectral sensitivities, resulting in ΔE < 1 and more accurate color reproduction than HDTV cameras [3]. Although the camera used is a specialized type, its spectral sensitivities can be considered typical because UHDTV cameras are designed to be oriented to an accurate color reproduction. Due to the accuracy of color reproduction, higher color rendering properties are required in lighting for UHDTV production. The transition of lighting from conventional lighting using incandescent lamps and fluorescent lamps to white light-emitting diode (LED) lighting is progressing owing to their high luminous efficacy concurrently with the transition from HDTV to UHDTV [4,5]. Although white LED lighting is expected to be implemented in television studios and sports facilities, appropriate indices for color rendition evaluating white LED lighting in UHDTV production are still unknown.
The color-rendering index (CRI) [6], which is defined by International Commission on Illumination (CIE), is a quantitative measure of the ability of a light source to reveal the colors of various objects in comparison to a black body radiator or daylight. The standard CRI values (R1–R8) are derived from CIE 1974 test-color samples consisting of eight low-saturation colors and six special CRI values (R9–R14) derived from the realistic colors red, yellow, green, blue, Caucasian skin color, and green foliage, respectively. The general CRI value (CIE Ra) is derived as the arithmetical mean of the eight standard CRI values (R1–R8). The recommended Ra value of a light source considered for TV production is often applied for lighting of sports facilities. The Japanese Industrial Standards (JIS) recommend Ra of 80 or higher for lighting during sporting events intended for HDTV production [7], which was decided using subjective evaluation with an actual HDTV system [8, 9]. The International Commission on Illumination (CIE) also recommends Ra > 80. Some international sports associations proposed higher recommendations for the color rendering properties of light sources to be used in HDTV production. For instance, the International Olympic Committee (IOC) has stated that lighting for field of play should approximate a television studio environment and all lamps shall have a correlated color temperature (CCT) of 5600 K and Ra ≥ 90 [10], and the International Basketball Federation (FIBA) has recommended Ra > 90 for lighting in stadiums used for international indoor basketball games [11]. Similarly, the International Association of Athletics Federations has also recommended Ra > 90 for lighting used for outdoor track and field events [12].
Although Ra is widely used internationally for evaluating illuminants, it is reported that highly saturated colors cannot be properly evaluated in terms of Ra under white LED lighting [13,14]. In order to overcome the defect of Ra and to evaluate white LED lighting properly, numerous new methods for evaluating color rendition have been proposed. The color quality scale (CQS) was developed by National Institute of Standards and Technology (NIST) where 15 saturated Munsell color samples are used [15]. The technical memorandum (TM-30) was developed by the Illuminating Engineering Society (IES) [14] where 99 color samples including saturated colors are used. Although many methods for evaluating color rendition have been developed to overcome the defect of Ra, none have yet been standardized. The television lighting consistency indices (TLCI), represented by Qa, was developed by the European Broadcasting Union (EBU) [16] where the ColorChecker was chosen for color samples. It is appropriate for evaluating HDTV systems, because Qa is calculated with TLCI model using the EBU camera spectral sensitivities based upon the average of the measurements of several HDTV cameras. Recently, the spectral similarity index (SSI) was proposed by the Academy of Motion Picture Arts and Sciences (AMPAS) [17] based upon the similarity of a spectrum to a reference spectrum that eliminates the need for any assumption of a specific observer or camera spectral sensitivity.
Meanwhile, the special CRI R9 is used to complement Ra until a new method for evaluating color rendition is established in the solid-state industry [18]. In the USA, grades of LED lighting are represented with a combination of Ra and R9 [19–21]. The Department of Energy has set a regulation of Ra ≥ 80 and R9 > 0 [22]. The California Energy Commission defined Ra ≥ 90 and R9 > 50 as regulation values of the bulb-type LED lamp [23]. The performance and quality requirement guidelines of LED lighting in Japan uses a combination of Ra and R9 that follows the one used in the U.S.A [24]. Many manufacturers specify the R9 value of the lighting device [25]. The combination of Ra and R9 is used for the regulations because the R9 value of a white LED tends to be significantly lower in spite of its high Ra value [26]. In some handbooks, it is stated that light sources with low values for R9 are less likely to be accepted for general illumination and it is recommended to check R9 in addition to Ra [27]. Moreover red objects have the maximum influence in general subjective evaluations [28], and the R9 value has a strong correlation with the appearance of the skin color of humans [29]. The reason for this correlation is considered to be that the trend of the spectral reflectance curve of R9 in the long wavelength area of λ > 580 nm is similar to that of human skin. Although R13, which was originally established for evaluating light skin tone, should be able to be used for this evaluation, its spectral reflectance curve is different from that of actual human skin. Therefore, R9 is an important index for evaluating color reproduction of human skin. Although the combination of Ra and R9 is not a perfect solution, it is recognized as a practical solution in the lighting industry to prevent further mismatch with conventional Ra [30].
Our goal is selecting appropriate indices for color rendition and determining their recommended values for white LED lighting in UHDTV. We conducted three experiments. In section 2, four types of white LED lighting and the types selected for our experiments, that will be used mainly in UHDTV production, were discussed. In section 3, the color-reproduction accuracies of UHDTV and HDTV cameras under white LED lighting were examined and it is confirmed that UHDTV cameras were more accurate in reproducing color than a HDTV camera even under white LED lighting. In section 4, the color differences reproduced by UHDTV and HDTV television systems between under a Xenon lamp simulating daylight and the four different white LED lighting using 24 color patches were evaluated, and it is revealed that a UHDTV camera is more sensitive to the color-rendering properties of the light source because of its color-reproduction accuracy. In section 5, a subjective evaluation using UHDTV system and various white LED lighting and reference lighting conditions was conducted and validities of CRIs (Ra, R9-14), CQS, TM-30, TLCI Qa, and SSI were examined. The appropriate indices for color rendition were selected and their recommended values were determined for white LED lighting in UHDTV production.
2. Types of white LED lighting for UHDTV production
Generally, white LED lighting has inferior color-rendering than incandescent lamps because of its non-smooth spectral power distribution. To overcome this drawback, white LED lighting has advanced in recent years. Because a fundamental trade-off exists between the color-rendering property and the luminous efficacy of lighting, various white LEDs have been developed to date and are categorized into four types, as outlined in Table 1. Type 4 white LEDs are mainly used for stage lighting to change color of cycloramas, and have poor color rendition and luminous efficacy because of the narrow bandwidth of each of the three component emission spectra and the low photoelectric conversion efficiency of green LEDs. Therefore, we excluded Type 4 white LEDs in this study. Types 1, 2, and 3, referred to as phosphor-converted white LED lighting, are considered to be the mainstream of white LED lighting. Among them, Type 1 is currently most used in luminous-efficacy-oriented purposes such as outdoor lighting, but their color rendering is inferior. Type 3 is the most effective at improving the color rendering properties of the light source; however, their luminous efficacies are low and the use of this type is limited to the case of evaluating accurate color, such as in museum or hospital lighting. Since it is difficult to make high power floodlight for television studio and sports facilities with Type 3, we also excluded Type 3 white LED lighting in this study. In contrast, Type 2 white LED lighting can obtain a higher luminous efficacy than Type 3 white LED lighting and have better color-rendering properties than Type 1 white LED lighting [31]. Consequently, Type 2 white LED lighting is expected to be widely used for television studios and sports facilities in the future. In this study, we focus on the Type 1 and 2 white LED lighting.
Table 1. Color rendering properties and luminous efficacy for four different types of white LED lighting.
3. Experiment I: Color reproduction accuracy of UHDTV and HDTV camera
In order to confirm the color reproduction accuracy of a UHDTV camera, we evaluated the color reproduction accuracy of UHDTV and HDTV cameras under white LED lighting with different color rendering properties.
3.1 Method
3.1.1 Illuminant
Four white LED lights with different spectral characteristics and a Xenon lamp that simulated daylight were used as illuminants. Figure 1 shows the spectral power distributions of the different lights. i: Type 1 LED – Efficacy oriented LED (Ra = 68) designed with a focus on luminous efficacy; ii: Type 1 LED – High-saturation color-rendering LED (Ra = 88) featuring high-chromatic-saturation color-rendering; iii: Type 1 LED – Skin-color-rendering LED (Ra = 94); iv: Type 2 LED – High-fidelity color-rendering LED (Ra = 92) featuring high-fidelity color-rendering in comparison to the CIE reference illuminant; and v: Xenon lamp (Ra = 96) that simulates daylight. The correlated color temperature (CCT) of the light sources was approximately 5,000 K. Table 2 shows the characteristics of the 5 light sources.
Table 2. Characteristics of 5 light sources used in Experiment I and II. (1,000 Duv equivalent to 1 duv)
3.1.2 Camera
Figure 2 shows the detailed method of Experiment I. A UHDTV camera prototype, fabricated by NHK [3,32], and a commercially available HDTV camera (HDL-45A, Ikegami, Japan) were tested. A spectroradiometer (SR-UL1, Topcon, Japan) was used to obtain the reference CIE tristimulus values of XYZ, which were used to calculate the color difference of the camera ΔEi (i = 1-24). Considering that a fixed linear matrix is usually used for television production, fixed linear matrices were applied for both cameras. They were optimized under the Xenon lamp (SOLAX XC-500, Seric, Japan) using a color chart (ColorChecker Classic, X-Rite, USA) by minimizing the sum of squares of the color difference ΔEi of the 24 color patches. To evaluate color difference ΔEi, we used ΔEab and ΔE00. Because the optimized linear matrices differ slightly between the cases of applying color difference ΔEab and ΔE00, we applied optimized linear matrices for evaluating the individual color differences. Evaluating the color difference ΔEi with the fixed linear matrix applied, the color differences were calculated by optimizing the gains of red, green, and blue channels so as to minimize the color difference ΔEi for each light type. Finally mean and maximum values of ΔEi were derived.
3.2 Results
Table 3 presents the mean and maximum values of color differences ΔEi using ΔEab and ΔE00 obtained in Experiment I. The mean color difference for the UHDTV camera is < 1 for all five light sources, whereas that for the HDTV camera ranges from 1.6 to 3 for the case of using color difference ΔEab. This suggests that color reproduction with UHDTV cameras is more accurate than that with HDTV cameras, even under white LED lighting. UHDTV cameras
Table 3. Color differences exhibited by UHDTV and HDTV cameras obtained in Experiment I.
4. Experiment II: Color differences reproduced by television system under Xenon lamp simulating daylight and white LED lighting
Next, in order to examine the effects of the color-rendering properties of lighting, we calculated the color differences reproduced by UHDTV and HDTV television systems under daylight and the four different white LED lighting.
4.1 Method
Figure 3 shows the detailed method of Experiment II. Color differences ΔEi between the reproduced colors of the 24 color patches captured under the Xenon lamp simulating daylight and under the four white LED lights were calculated. To calculate color difference, we used ΔEab and ΔE00 similarly to Experiment I. The linear matrix obtained in Experiment I was applied. The white balance was adjusted by tuning the gain for each light type. The Rec. 709 nonlinear optoelectronic transfer function (OETF) was applied to each camera. Two functions were used for the electro-optical transfer function (EOTF): a 2.4-power function, in compliance with Recommendation ITU-R BT.1886 [33], and the inverse of Rec. 709 OETF.
4.2 Results
Tables 4 and 5 present the mean and maximum differences between the color rendition of the individual color patches captured under the Xenon lamp and each white LED lighting when the EOTFs were used. In general, the mean and maximum color differences were greater for the UHDTV system than for the HDTV system, which indicates that the UHDTV camera is more sensitive to the color rendering properties of the light source because of its color reproduction accuracy. This suggests that improved color-rendering properties of lighting are required for UHDTV production.
Table 4. Mean and maximum values of the differences between the reproduced colors of the 24 color patches captured under the daylight and each white LED lighting for 2.4-power EOTF obtained in Experiment II.
Table 5. Mean and maximum values of the differences between the reproduced colors of the 24 color patches captured under the daylight and each white LED lighting when using the inverse of Rec. 709 OETF obtained in Experiment II.
The color difference caused by the difference in color rendering properties varies depending on the camera. It is thought that it depends on the spectral sensitivity of the camera. The spectral sensitivities of the UHDTV camera as shown in Fig. 10 of Ref [3]. are very close to the ideal spectral sensitivities of UHDTV, which results in its high color reproducibility. On the other hand, the spectral sensitivities of the general HDTV camera as shown in Fig. 9 of Ref [34]. are very different from the ideal spectral sensitivities of a HDTV camera (see Fig. 1 of Ref [3].), which results in its low color reproducibility. It is thought that this difference in color reproducibility appears in the results of Experiments I and II.
5. Experiment III: Subjective evaluation to select appropriate indices for color rendition and determine recommended values for UHDTV production
We conducted a subjective evaluation using a UHDTV system and various white LED lighting and reference lighting conditions and selected the appropriate indices for color rendition from among various methods of evaluating color rendition of Ra, special CRI values (R9–R14), TLCI Qa, CQS Qa, CQS Qf, TM-30 Rf, and SSI, and determined their recommended values for white LED lighting in UHDTV production.
5.1 Method
5.1.1 Participants
Thirty-four non-expert participants (18 females and 16 males; age range = 20–51 years; mean age = 36.1 years) with normal vision (in compliance with Recommendation ITU-R BT.500) [35] participated in the subjective experiment.
5.1.2 Apparatus
To create different lighting conditions, a special booth consisting of LEDs with 24 different peak wavelengths was used. Three levels of CCT were employed: 3,000 K (representing studio lighting), 5,600 K [36, 37] (stadium lighting), and 6,500 K (corresponding to the reference white light of a television system). For each CCT, we configured five lighting conditions, including four simulated white LED lighting ones and one as a reference lighting condition, with different color rendering indices values. Four simulated white LED lighting were selected considering their popularity and commercial availability. The reference lighting condition at 3,000 K was designed to approximate blackbody radiation and the other two reference lighting conditions were designed to approximate the CIE daylight illuminant. Figure 4 shows the spectral power distributions of the 15 lighting conditions and Table 6 shows a summary of their colorimetric characteristics. We also evaluated these with corresponding various methods of evaluating color rendition (Ra, R9–R14, TLCI Qa, TLCI Qa´, CQS Qa, CQS Qf, TM-30 Rf, and SSI) of each lighting condition. For calculating TLCI Qa, we applied the model of methods of Tech 3355 [38]. Although TLCI Qa defined in Tech 3355 is calculated based on the spectral sensitivities of a typical HDTV camera, that is different from that of a UHDTV camera, we redefined TLCI Qa´ for a UHDTV using the ideal spectral sensitivities of a UHDTV camera. The UHDTV camera from Experiment I was used to capture the evaluation images. Figure 5 shows our experimental setup. A reference image was captured under reference lighting condition, and 4 comparison images were captured under four simulated white LED lighting conditions for each CCT. A linear matrix optimized under simulated daylight at a CCT of 5,600 K was used because of its relatively balanced spectral power distribution at short and long wavelengths.
Fig. 4 Spectral power distributions of the lighting conditions used in Experiment III at CCTs of (a) 3,000 K, (b) 5,600 K, and (c) 6,500 K.
Table 6. Characteristics of lighting conditions used in Experiment III. (1,000 Duv equivalent to 1 duv)
The white balance was adjusted under each lighting condition using the color chart (ColorChecker Classic, X-Rite, USA), and the gamma correction defined by Rec. 2020/Rec. 709 was applied. The images were displayed on a 50-inch laser-backlit 4K LCD monitor with 98% coverage of the Rec. 2020 color space with a 2.4-power EOTF in compliance with Recommendation ITU-R BT.1886 being applied [39].
5.1.3 Materials
Two groups of objects were prepared. The “Sports” group included uniforms, samples of artificial lawn, natural foliage, synthetic clay, national flags, a soccer ball, and color charts (ColorChecker Classic and CIE 1974 test-color samples); the colors were mostly within the Rec. 709 color gamut. The “Flowers” group comprised a wide range of colors and included fresh and artificial flowers, fruits, toy blocks, wine glasses, bottles, and color charts. In the subjective evaluation, the color charts used in the Sports group were evaluated in another group called Charts (see Fig. 6).
Fig. 6 Evaluation images used in Experiment III and their chromaticity distributions. Sports group: (a) Evaluation image, (b) chromaticity distribution. Flower group: (c) Evaluation image, (d) chromaticity distribution.
5.1.4 Procedure
The experiments were conducted in dim surroundings. The display brightness was adjusted to ensure that the white patch of the ColorChecker Classic chart had a luminance of 150 cd/m2. The luminance level of the wall behind the monitor was set to a range of 2–10 cd/m2. This setup was used to ensure that the ratio of the peak luminance of the screen image to that of the monitor background remained within a range of approximately 1–5%. The viewing distance was set to 1,900 mm, which was three times the screen height.
The reference image and one of four comparison images captured under the same CCT lighting conditions were displayed side by side, divided by a gray border. The order of the comparison images to be displayed was randomized. The participants were able to move the gray border horizontally using a dial, as shown in Fig. 7. Four comparison images were evaluated for each of the three groups (i.e., Sports, Flowers, and Charts as a subgroup of Sports) at three CCTs (3,000, 5,600, and 6,500 K), for 36 trials per participant. A session lasted approximately 30 min and included 12 trials. There was a 30 min break between sessions.
Fig. 7 Photograph of subjective evaluation of Experiment III. A participant evaluates a reference image (left side) and a comparison image (right side) displayed side by side controlling the gray border.
The participants evaluated each image using the degradation category rating (DCR) method with a five-stage degradation scale, as shown in Table 7. In the DCR method, scores of 4.5, 3.5, and 2.5 are quality standards that define the detection limit, acceptability limit, and bearable limit, respectively [40, 41]. The raw experimental data obtained by the DCR method with a five-stage degradation scale are not equal-interval scale values (equal-interval scale means equal distances anywhere along the scale representing the same significance) but just category scores. Thus, we converted raw category scores to equal-interval scale values by adopting the method of successive categories [42].
Table 7. Five-stage degradation scale of DCR method used in Experiment III.
Before the experiment was conducted, we examined the correlation between the DCR method and the Thurstone method using the reference image and four comparison images with 22 participants with 5,600 K CCT conditions. As shown in Fig. 8, we confirmed a very high correlation (>0.9) between the two methods, and the high linearity of the five-stage degradation scale of the DCR method.
5.2 Results
Using the method of successive categories, the five category scores of 1 to 5 were respectively converted into equal-interval scale values of 0, 1.033, 1.834, 2.745, and 3.963. Figure 9 shows the average of the equal-interval scale values obtained using the DCR method for each CCT. The highest scores were obtained with the high-fidelity color rendering LED in every condition. Figures 10-12 show the comparison of equal-interval scale values versus indices values of various methods of evaluating color rendition (Ra, R9–R14, TLCI Qa, TLCI Qa´, CQS Qa, CQS Qf, TM-30 Rf, and SSI). In the comparison of the equal-interval scale values versus SSI, the two sizes of markers denote the two CCTs (3,000 K and 5,600 K) because the SSI is not defined for 6,500 K. Pearson product-moment correlation coefficients (r) were also shown in each figure (Pearson product-moment correlation coefficients can be applied to equal-interval scale values). These charts and correlation coefficients show that R9 correlates best with the equal-interval scale values. In addition, CQS Qa, and TM-30 Rf show high correlation with the equal-interval scale values and have a higher correlation than the conventional Ra. This suggests that the new method for evaluating color rendition perform better than Ra.
Fig. 9 Equal-interval scale values for the color differences in UHDTV imagery obtained in Experiment III for each CCTs of (a) 3,000 K, (b) 5,600 K, and (c) 6,500 K. Marker shapes denote three objects. Marker sizes denote three CCTs.
Fig. 10 Comparison of equal-interval scale values versus Ra obtained in Experiment III. The markers denote four white LED lighting conditions, three CCTs, and three objects as shown in Fig. 9.
Fig. 11 Comparison of equal-interval scale values versus six special CRI values of (a) R9, (b) R10, (c) R11, (d) R12, (e) R13, and (f) R14 obtained in Experiment III. The markers denote four white LED lighting conditions, three CCTs, and three objects as shown in Fig. 9.
Fig. 12 Comparison of equal-interval scale values versus indices values of new methods for evaluating color rendition of (a) TLCI Qa, (b) TLCI Qa´, (c) CQS Qa, (d) CQS Qf, (e) TM-30 Rf, and (f) SSI obtained in Experiment III. The markers denote four white LED lighting conditions, three CCTs, and three objects as shown in Fig. 9. In the comparison of the equal-interval scale values versus SSI, the two sizes of markers denote two CCTs (3,000 K and 5,600 K) because the SSI is not defined for 6,500 K.
6. Discussion
6.1 Appropriate indices for color rendition and their recommended values
Experiment III confirmed that R9 correlates best with the equal-interval scale values and that choosing R9 as an additional index is valid. Considering the compatibility with the existing index of Ra, it is practically appropriate to use the combination of Ra and R9. Although it is not a perfect solution, it seemed to be reasonable to use R9 complementing Ra until a new more appropriate index is standardized to prevent further mismatch with conventional Ra. In addition to this, CQS Qa and TM-30 Rf, which had high correlations with equal-interval scale values, can be also considered as probable candidates. Therefore, we derived the recommended value for the combination of Ra and R9 and CQS Qa and TM-30 Rf by applying the acceptability limit of DCR method. The acceptability limit of 3.5 in the DCR method becomes 2.289 equal-interval scale value converted by method of successive categories. To obtain the recommended value of each index, we plotted the distributions between each index values and the equal-interval scale values with 95% confidence intervals as shown in Fig. 13. The 95% confidence interval includes the population mean of the equal-interval scale values in the DCR method with 95% probability. If we adopt a condition that exceeds the acceptability limit including the 95% confidence interval, the population average of equal-interval scale value in the DCR method should almost always exceed the acceptability limit. In this figure, the lighting conditions satisfying the acceptability limit including 95% confidence intervals are indicated by the red line. It is sufficient to decide the recommended value so that only the lighting conditions indicated by red line are classified. When the combination of Ra and R9 is selected, the condition of Ra ≥ 90 and R9 ≥ 80 is considered appropriate. Although there is a distribution that does not satisfy the acceptability limit with the condition of Ra ≥ 90 alone, by adding a condition of R9 ≥ 80, the acceptability limit is satisfied. This condition is consistent with the conventional practice of determining the recommended CRI values in 10 increments [43]. When CQS Qa and TM-30 Rf are selected, the conditions of CQS Qa ≥ 93 or TM-30 Rf ≥ 90 are derived for classifying the lighting conditions shown as red lines in Fig. 13. In conclusion, it is considered that around 90 or more will be appropriate for CQS Qa and TM-30 Rf. At present, the combination of Ra and R9 is appropriate, and the conditions of Ra ≥ 90 and R9 ≥ 80 are recommended for white LED lighting in UHDTV production.
Fig. 13 The distributions between the equal-interval scale values and each index values of (a) Ra, (b) R9, (c) CQS Qa, and (d) TM-30 Rf with 95% confidence intervals. The markers denote four white LED lighting conditions, three CCTs, and three objects as shown in Fig. 9.
The subjective evaluation in Experiment III was conducted with limited lighting conditions using only four types of LED lighting that were selected based on practicality and commercial availability. Further research is required to evaluate other types of LEDs. However, even if other indexes than Ra and R9 were chosen, the effectiveness of the combination of Ra and R9 would remain taking into account the fact that the other indices do not correlate as well with the equal-interval scale values as R9, at least for the four types of LED lighting. It is advisable to use the combination of Ra and R9 while waiting for the establishment of a new single faultless index, rather than creating confusion by adopting new incomplete indices one after another.
This study involved only Japanese participants. We have not considered any impact on the results that may be due to racial differences, which we would like to research in the near future. However, because there are no adjustment items due to racial differences in the color difference formulae, we think that racial difference cannot be a major factor in the evaluation of color reproduction based on color difference.
7. Conclusion
We discussed appropriate indices for color rendition and their recommended values of white LED lighting for UHDTV production. We confirmed that the UHDTV camera was more accurate in reproducing color than a HDTV camera even under white LED lighting and confirmed that lighting with higher color-rendering properties is required for the UHDTV system because of its color-reproducing accuracy. We conducted a subjective evaluation to select the appropriate indices for color rendition and to determine their recommended values. From the correlation analysis between equal-interval scale values and various methods of evaluating color rendition (Ra, R9–R14, TLCI Qa, TLCI Qa´, CQS Qa, CQS Qf, TM-30 Rf, and SSI) for each lighting type, the highest correlation was achieved by R9. The combination of Ra and R9 was found to be practically appropriate and recommended values of Ra ≥ 90 and R9 ≥ 80 were obtained for white LED lighting in UHDTV production.
References and links
1. Recommendation ITU-R BT.2020–2, “Parameter values for UHDTV systems for production and international programme exchange” International Telecommunications Union, Geneva, 2015.
2. Recommendation ITU-R BT.709–6, “Parameter values for the HDTV standards for production and international programme exchange” International Telecommunications Union, Geneva, 2015.
3. K. Masaoka, Y. Nishida, T. Soeno, T. Yamashita, M. Sugawara, and A. Saita, “Designing camera spectral sensitivities for UHDTV,” SMPTE Mot. Imag. J. 123(8), 26–32 (2014). [CrossRef]
4. M. Wei and K. W. Houser, “Status of solid-state lighting based on entries to the 2010 US DOE next generation luminaire competition,” Leukos 8(4), 237–259 (2012).
5. J. B. Protzman and K. W. Houser, “LEDs for general illumination: The state of the science,” Leukos 3(2), 121–142 (2006).
6. CIE Publication 13.3, “Method of measuring and specifying colour rendering of light sources” International Commission on Illumination, Vienna, 1995.
7. JIS Z 9127:2008, “Recommendation for sports lighting” Japanese Standards Association, Tokyo, 2008.
8. Y. Tanabe, M. Shimizu, M. Fujimoto, and N. Hata, “A study of the colour rendering of lamps for HDTV systems” Lux Pacifica, E40–E44, 1993.
9. O. Kobayashi, T. Soda, Y. Hirashima, and Y. Tamura, “The evaluation of colour rendering of lamps with HDTV camera,” IESNA Annu. Conf., 1992.
10. International Olympic Committee, “Host city contract operational requirements: BRS annex 3 specifications on lighting” International Olympic Committee (2015).
11. FIBA Central Board, “Official basketball rules 2014 – Basketball equipment” in official basketball rules 2014, International Basketball Federation, Geneva, 2014.
12. IAAF Track and Field Facilities Manual Editorial Board, “Technical services” in IAAF Track and Field Facilities Manual 2008 Edition, International Association of Athletics Federations, Monaco, 2014.
13. K. Smet, W. R. Ryckaert, M. R. Pointer, G. Deconinck, and P. Hanselaer, “Correlation between color quality metric predictions and visual appreciation of light sources,” Opt. Express 19(9), 8151–8166 (2011). [CrossRef] [PubMed]
14. A. David, P. T. Fini, K. W. Houser, Y. Ohno, M. P. Royer, K. A. Smet, M. Wei, and L. Whitehead, “Development of the IES method for evaluating the color rendition of light sources,” Opt. Express 23(12), 15888–15906 (2015). [CrossRef] [PubMed]
15. W. Davis and Y. Ohno, “Color quality scale,” Opt. Eng. 49(3), 033602 (2010). [CrossRef]
16. R-137, “Television lighting consistency index-2012” FTV-LED, Geneva, November 2012.
17. J. Holm, T. Maier, P. Debevec, C. LeGendre, J. Pines, J. Erland, G. Joblove, S. Dyer, B. Sloan, J. di Gennaro, and D. Sherlocky, “A cinematographic spectral similarity index,” SMPTE 2016Annual Technical Conference.
18. Voice of the Lighting Industry, “Lighting europe position paper on color quality” Lighting Europe, 6 October 2014.
19. “CALiPER” Application Summary Report 22: LED MR16 Lamps, U.S. Department of Energy, June 2014.
20. “Bright tomorrow lighting competition (L Prize)” Competition Requirements for PAR 38 Halogen Replacement Lamps, U.S. Department of Energy, July 2013.
21. “Solid state lighting annex: Product quality and performance tiers” Efficient Electrical End-Use Equipment (4E), International Energy Agency, 17 December 2015.
22. “ENERGY STAR® program requirements for lamps (Light bulbs)” International ENERGY STAR® Program, August 2014.
23. Codes and Standards Enhancement (CASE) Initiative, “LED lamp quality” California Energy Commission, 2014.
24. “Performance request guidance in high-quality lighting for the LED light source” Japan Lighting Manufactures Association, 17 April 2013. (in Japanese)
25. Catalogs of Nichia Corporation, Citizen Electronics Co., Ltd.
26. Codes and Standards Enhancement (CASE) Initiative for PY 2014: Title 20 Standards Development, “Small diameter directional lamps” California Energy Commission, Sacramento, 2014.
27. Illuminating Engineering Society, The Lighting Handbook: Reference and Application, 10th ed. (Illuminating Engineering, 2011).
28. S. Jost-Boissard, M. Fontoynont, and J. Blanc-Gonnet, “Perceived lighting quality of LED sources for the presentation of fruit and vegetables,” J. Mod. Opt. 56(13), 1420–1432 (2009). [CrossRef]
29. Building Technology Program, “LED color characteristics” Solid-State Lighting Technology Fact Sheet, U.S. Department of Energy, PNNL-SA-84900, January 2012.
30. Ian Ashdown, “Thoughts on color rendering” Lighting Analysts Inc. https://www.researchgate.net/publication/273763307_Thoughts_on_Color_Rendering
31. “Problems related to LED lighting fixtures and study of standardization of its construction” Japan Electrical Construction Association, December 2015 (in Japanese).
32. T. Yamashita, R. Funatsu, T. Yanagi, K. Mitani, Y. Nojiri, and T. Yoshida, “A camera system using three 33-m-pixel CMOS image sensors for UHDTV2,” SMPTE Mot. Imag. J. 120(8), 24–31 (2011). [CrossRef]
33. Recommendation ITU-R BT.1886-0, “Reference electro-optical transfer function for flat panel displays used in HDTV studio production” International Telecommunications Union, Geneva, 2011.
34. P. Bohler, J. Emmett, and A. Roberts, “Toward a “standard” television camera color model,” SMPTE Mot. Imag. J. 122(3), 30–36 (2013). [CrossRef]
35. Recommendation ITU-R BT.500-13, “Methodology for the subjective assessment of the quality of television pictures” International Telecommunications Union, Geneva, 2012.
36. C. Clay, “Lighting for sports stadiums,” Lighting J. 70, 26–29 (2005).
37. A. Kaufman and D. Sauter, “Problems of lighting in color television outdoor broadcasts,” J. SMPTE 83(1), 20–26 (1974). [CrossRef]
38. Tech 3355, “Method for the assessment of the colorimetric properties of luminaires the television lighting consistency index (TLCI-2012)” FTV-LED, Geneva, April 2013.
39. S. Maeda, N. Okimoto, H. Yasui, A. Heishi, E. Niikura, K. Minami, Y. Nishida, and Y. Kusakabe, “RGB laser-backlight LCD”, Proc. ITE annual convention, 32A–1 2015.
40. T. Hasegawa, M. Inoue, and T. Mitsuhashi, “Subjective assessment of TV picture quality and data processing,” J. Inst. Television Eng. Jpn. 37, 1040–1050 (1983). [CrossRef]
41. J. Okamoto and T. Hayashi, “Latest trends in image media quality assessment technologies,” IEICE Fundam. Rev. 6(4), 276–284 (2013). [CrossRef]
42. M. Saffir, “A comparative study of scales constructed by three psychophysical methods,” Psychometrika 2(3), 179–198 (1937). [CrossRef]
43. CIE Publication 29.2, “Guide on interior lighting” International Commission on Illumination, Vienna, 1986.
