A multiprimary-color display provides high fidelity and wide-gamut color reproduction, as well as power savings and other advantages. New methods will be required for evaluating multiprimary displays, particularly in terms of color gamut, color accuracy, and smooth tone reproduction.
by Masahiro Yamaguchi
MULTIPRIMARY-color displays have begun to enter the consumer market. Their advantages include a wide color gamut, power savings, and suitability for high-fidelity color reproduction. How will these displays impact professional applications and what new aspects are needed for evaluating color displays so as to best incorporate the advantages of multi-primary devices?
Multiprimary-color displays
Conventional displays reproduce colors based on the additive mixture of R (red), G (green), and B (blue) primary colors. The color gamut (range of colors reproducible by displays) is limited to the triangle spanned by the RGB primary colors. To realize wider-gamut displays, two different methodologies have been attempted: (1) more-saturated RGB primaries and (2) the use of more than three primary colors – the multiprimary approach. Wide-gamut displays such as LED-backlit liquid-crystal displays (LCDs)1 and laser displays have been extensively commercialized based on the first approach. Multiprimary-color displays have more recently been investigated. Examples include multi-projection displays,2,3 a projec-tor with a multiprimary color-filter wheel,4,5 a color pixel generated with a diffraction grating,6 a time-sequential backlight using a multiprimary LED light source,1,7 and an LCD with a multiprimary color-filter array.8,10
Figure 1 shows a comparison of color gamuts for a conventional standard RGB (sRGB) display, an LED-backlight wide-gamut display,1 and a six-primary-color display.3,11 The gamut of a six-primary-color display covers almost 100% of real-world object colors, while a conventional display gamut, such as sRGB or ITU-R BT.709, covers only 80%. This is one of the main reasons that the multiprimary approach is advantageous in high-fidelity color reproduction. The details of the evaluation of the size of the gamut will be explained further on.
Advantages of Multiprimary-Color Displays
As shown in Fig. 1, the gamut of multi-primary-color displays efficiently covers real-world object colors. The design of the gamut shape becomes more flexible than that of wide-gamut RGB displays, in which the chromaticity coordinates of the G primary are often moved to shorter wavelengths so that the gamut covers the cyan region, but the yellow region is excised. In multiprimary displays, cyan and/or yellow can be added, and the yellow-green-cyan region can be broadly covered. In the three-primary case, narrow-band primary colors, especially R primaries of longer wavelengths and B primaries of shorter wavelengths, are used, but the sensitivity of human vision is lower at such wavelengths so that the overall efficiency is decreased.
Fig. 1: The graph represents the color gamuts of an ITU-R BT.709 (or sRGB) display, an LED-backlight wide-gamut display,1 and a six-primary-color display.3,11 The dot distribution represents the gamut of real-world object colors.
The advantage of multiprimary-color displays is not only wide-gamut color reproduction. A multiprimary-color display also enables the reproduction of color with spectral approximation,12 in which the spectral difference between the original and reproduced colors is minimized in addition to the reproduction of colorimetric tristimulus values. Although conventional color reproduction is based on the color-matching function (CMF) of a standard observer defined in a CIE standard, such as CIE 1931 XYZ, the spectral sensitivities of real observers are slightly different from the standard CMF. This leads to a color mismatch between the real object and the image reproduced on a monitor, even though the color difference may be very small. Spectral approximation solves this problem. Unlike with a conventional colorimetric (metameric) match, the color matching can be achieved even if the observer variation of CMF is not ignored. Metamerism involves matching the apparent colors of objects with different spectral power distributions. "Spectral displays" are valuable in applications that require highly accurate color matching, such as soft-proofing (reviewing a print job on a computer monitor rather than on paper).
It has been recently publicized that power savings is an important feature of multi-primary displays.10 Adding a yellow primary is effective for this purpose, and the design of color conversion is a key issue in reducing power consumption. A power savings of 20% has been achieved in a four-primary-color display10 compared with a conventional RGB display. Even though some additional cost is required in color signal processing, the complexity of the color-processing engine is not significantly higher than that of a standard consumer television set.
Multiprimary-color conversion introduces a new flexibility in the design of color displays. The viewing-angle performance, which is a critical issue in LCDs, can be improved in a five-primary-color display by modifying the multiprimary-color conversion technique.9 Also, multiprimary-color subpixels can be used to enhance the visual resolution by using subpixel rendering.
Signal Processing in Multiprimary- Color Displays
The multiprimary-color signal is generated from imaging tristimulus values or multi-spectral data, called multiprimary-color conversion, as shown in Fig. 2, similar to the color decomposition for multicolor printers. For colorimetric color reproduction, three-dimensional tristimulus values (such as XYZ or YCbCr) are transformed to M-dimensional multiprimary-color values if the number of primary colors is M. This methodology involves a degree of freedom; plural combinations of multiprimary color values can reproduce a certain color. There have been various methods proposed for multiprimary-color conversion.
Fig. 2: The system configuration of a multiprimary-color display for colorimetric color reproduction is shown above and for spectral color reproduction, below.
In the matrix-switching (MS) method,13 the polyhedral color gamut spanned by M-primary colors is divided into pyramids, and a linear color conversion is performed in each pyramid. Another approach is based on linear programming.14 Linear programming enables the incorporation of various criteria for selecting multiprimary-color-value combinations. For example, as the MS method sometimes induces contour-like artifacts at the boundary of pyramids, a constraint can be introduced to improve the smooth tone reproduction without such artifacts. In this case, multiprimary signal values are decomposed into visible and invisible [metameric black (MB)] components. The visible components are uniquely solved, where the MB components are determined such that the multiprimary signal values change smoothly if the change of tristimulus values is small.14 Linear programming is also a powerful technique for incorporating extra features to a multiprimary-color display, such as power saving and the improvement of viewing-angle performance, as mentioned in the previous section.
For the spectral color reproduction explained previously, multiprimary device signal values are directly calculated from multispectral captured images. In the proposed conversion method for spectral displays, called the spectral approximation method,12 the spectral error is minimized under the constraint in such a way that a colorimetric match is attained for the standard observer.
Industrial and Medical Applications
High-fidelity color reproduction is required in industrial and medical applications such as printing, textiles, industrial design, digital archives of artworks, and true-color medical images. However, it is difficult to capture high-fidelity color with RGB-based color-imaging systems. An approach to overcoming such limitations is going beyond RGB – namely, adopting a spectrum-based system.11,15 In the spectrum-based color reproduction system, the color signal captured by a multiband camera is used to derive the spectral image, and the output color is rendered using the spectral image and the spectrum of illuminant. A multiband camera with more than three-channels, i.e., a multispectral camera, is useful for capturing spectral imagery with high accuracy. A multispectral imaging system enables both image capture with accurate color under various different illuminants and the reproduc-tion of very natural and realistic images. The conventional sRGB gamut has been shown to be insufficient for printing, textiles, and artworks. Based on a color-gamut analysis of color-printer inks, textile samples, and oil paints, wide-gamut displays and wide-gamut image capture should be more appropriate for these applications.
High-fidelity color is important in medical color displays, especially for telemedicine applications such as teledermatology, telepediatrics, and home-care telemedicine. The effectiveness of multiprimary displays in medical applications has not yet been proven, but some experimental results show the potential of using wide-gamut and multiprimary displays in these scenarios.
For example, a telemedicine experiment was carried out in order to investigate whether dermatologists perceived the identical color from a reproduced image and the original skin, and how the color reproducibility influenced the skin lesion diagnosis.17 For the preliminary test of the experiment, a conventional RGB CRT display with proper calibration was used, but the visual disagreement between the reproduced color and the original object was significant, despite the fact that the colorimetric accuracy was considerably high. The disagreement originated from the observer metamerism effect; the CMFs of real observers are slightly different from the standard CMF used in color measurement. This result does not necessarily mean that a three-primary-color display is inadequate for skin-color reproduction, but it does show the potential benefits of spectral displays. The experimental results showed that dermatologists perceived almost the same color in the case of the multispectral system, while perceived colors shifted to the red in the observation with an RGB system. In addition, there were oversights of erythema (skin redness) found in the RGB example, but no such oversights with the multispectral system. These results suggest the advantages of using the total multispectral system for accurate skin-color reproduction.
Another area of possible application is the video capture of surgery for case archives, conferences, education, and telesurgery with support from an expert surgeon at a remote site. The selection of the resection area is sometimes decided by faint color differences, which means natural-color reproduction is significantly important. An experiment was conducted18 in which seven examples of hand operations were captured by a six-band video camera and visually evaluated by surgeons and derma-tologists. Figure 3 shows the distribution of colors obtained from the six-band camera. The colors of tissues ranged from red to yellowish-white. The color of blood exceeded the sRGB gamut. It is probably advisable to employ a wide-gamut display, particularly in the deep red region, for reproducing blood color. In the experiment reported in Ref. 18, however, a flat-panel LCD with a normal color gamut was used, since another device suitable for this experiment was not available.
Fig. 3: Color distribution of tissues involved in medical operations on hands was captured by a six-band camera (CIE xy-chromaticity coordinates).
Evaluation of Multiprimary Displays
Evaluation methods for multiprimary-color displays are in some instances different from those standardized for conventional RGB displays. The color gamut of a non-standard display is often indicated with the NTSC (National Television Systems Committee) ratio, while the 1953 NTSC standard is not used in the current television systems. In fact, 3-D color space is advisable in order for a user to see the details of gamut size. Uniform chromaticity space, such as CIELAB color space, is suitable, and the volume of gamut in CIELAB color space can be an indicator of gamut size. Figure 4 shows the comparison of gamut shape in CIELAB space. It is visualized with the gamut expanded especially in bright orange, cyan, dark red, and dark purple regions in the six-primary-color display19,20 (see also "The NTSC Color Triangle Is Obsolete, But No One Seems to Know" in the May 2006 issue of Information Display.)
In addition, it is expected that a display gamut will cover the colors of real-world objects. The surface color data published by Pointer21have been sometimes used to evaluate wide-gamut displays. (Pointer measured real-object colors, including not only natural objects such as butterflies and flowers, but man-made ones such as paint samples and fabrics.22) However, the Pointer gamut is not necessarily sufficient for representing real-world natural colors. The colors calculated from the data in the Standard Object Color Spectra (SOCS) database in the ISO TR 16066:2003 standard can be compared with the Pointer gamut, and about 3800 among 50,000 samples are from the Pointer gamut. Thus, the SOCS database was integrated with the Pointer gamut for more representative reference. Figure 4 shows a plot of the merged Pointer + SOCS gamut boundary at different luminance on xy-chromaticity co-ordinates (fluorescent colors are excluded). It can be seen that the six-primary-color display covers the Pointer + SOCS gamut. Moreover, the outer boundary of the merged Pointer + SOCS is also designated in 3-D color space in Fig. 4. The coverage of object color is also a reasonable indicator of the gamut size. Table 1 shows an example of evaluation. The gamut volume of a six-primary-color display is about 1.8 times larger than a conventional RGB display and natural color coverage is larger than 99%.
Fig. 4: The color gamut of conventional RGB, six-primary colors, and Pointer + SOCS on constant lightness planes are shown at L* = 30, 50, and 90.
Color Accuracy
The accuracy of reproduced color can be evaluated by the difference between the target color and the reproduced color obtained by colorimetric measurement, as in the case of a conventional RGB display. However, it should be considered that color-reproduction errors originate from three sources: (1) the display color gamut, (2) the device characterization error, and (3) the color-conversion method. As for source (1), how is it possible to compare the color-reproduction accuracy of displays with different gamut shapes? How to select test colors? A possible way is to define the test colors within the Pointer + SOCS gamut and to derive the accuracy for the test colors inside the display gamut as well as for all the colors. Regarding the source of the color-reproduction error,3 it is usually insignificant in conventional RGB displays, but cannot be ignored in multiprimary displays because the multiprimary-color conversion involves a somewhat more complicated process than the RGB system. For example, a certain color can be reproduced by a different combination of multiprimary device signals. Thus, the reproducibility of a specific color depends on the multiprimary-color conversion method. While the device characterization can be performed by measuring the chroma-ticity points of primary colors, the tone-reproduction curve, and the background color, the accuracy of multiprimary-color conversion should be also evaluated.
Multiprimary-color conversion is sometimes an internal process in the commercial product, as in Fig. 2, and the input signal is in XYZ or YCC color space. In this case, it is impossible to control the primary-color signal independently. From the user's perspective, the conventional device used to make an ICC profile (a set of color-management data for a particular device, as established by the International Color Consortium) is not suitable for multiprimary-color displays. Accurate and efficient ways of characterizing multiprimary displays on the user side is an important issue.
Smoothness of Tone Reproduction
Since the 3-to-M multiprimary-color conversion involves a degree of freedom, the device signal may vary largely even for a slight change of target color. Ideally, an observer cannot perceive such variation. However, when a smooth tonal gradation is reproduced on a multiprimary-color display, contour-like artifacts sometimes appear.23 These result from characterization errors of the display and the variation in the color-matching functions of observers. For example, even if the artifact is not obvious when observed from the front of the display, it appears quite clearly at large observation angles because of the change in the spectral composition of reproduced light. The appearance of the contour-like artifacts depends on the multiprimary-color conversion. To reduce the artifact, the conversion method should ensure the continuity of device signal values for the continuous change of tristimulus values. The reproduction of smooth tonal gradation is one of the important issues in the image quality of a multiprimary display.
The number of reproducible colors in a multiprimary display is 2MB in principle, where M and B are the number of primary colors and bit depth of each channel, respectively. For example, in the case of a six-primary-color display with 8 bits/channel, the number becomes 2.8 x 1015 colors. However, the quantized points are not uniform in 3-D color space. In addition, if a specific method for multiprimary-color conversion is adopted, the number of colors used for reproduction is much smaller; for example, 3.4 x 109 colors in the case of an MS color-conversion method. A method for dealing with the unused signal values has yet to be developed.
Summary
A multiprimary-color display is beneficial for high-fidelity and wide-gamut-color reproduction, but also provides other advantages, such as power savings, spectral display, etc. The integrated design of the device with the multiprimary-color conversion is key to realizing the above advantages. Some modified methods are required for evaluating multiprimary-color displays, particularly with regard to color gamut, color accuracy, and smooth tone reproduction. The standardization of evaluation methods suitable for multiprimary-color displays will be an important contribution to this field.
Acknowledgment
This article includes results from the Natural Vision project supported by NICT (National Institute of Information and Communications Technology, Japan).
References
1H. Sugiura, S. Kagawa, H. Kaneko, M. Ozawa, H. Tanizoe, T. Kimura, and H. Ueno, "Wide color gamut displays using LED backlight- signal processing, circuits, color-calibration system, and multi-primaries," ICIP '05 II, 9–12 (2005).
2R. Yajima, M. Kanazawa, and S. Sakaida, "Wide-color-gamut system," Proc. ITE '94 Annual Convention, 355–356 (1994) (in Japanese).
3T. Ajito, T. Obi, M. Yamaguchi, and N. Ohyama, "Expanded color gamut reproduced by six-primary projection display," Proc. SPIE 3954, 130–137 (2000).
4S. Roth, I. Ben-David, M. Ben-Chorin, D. Eliav, and O. Bben-David, "Wide-gamut high-brightness multiple-primaries single-panel projection displays," SID Symposium Digest 34, 118–121 (2003).
5M-C. Kim, Y-C. Shin, Y-R. Song, S-J. Lee, and I-D. Kim, "Wide-gamut multi-primary display for HDTV," Proc. CGIV '04, 248–253 (2004).
6T. Ajito, T. Obi, M. Yamaguchi, and N. Ohyama, "Multiprimary color display for liquid crystal display projectors using diffraction grating,"Opt. Eng. 38, No. 11, 1883–1888 (1999).
7S. Komura, I. Hiyama, and N. Ohyama, "Four-Primary-Color LCD for Natural Vision," Information Display 8/03, 18–21 (2003).
8E. Chino, K. Tajiri, H. Kawakami, H. Ohira, K. Kamijo, H. Kaneko, S. Kato, Y. Ozawa, T. Kurumisawa, and K. Inoue, "Development of Wide-Color-Gamut Mobile Displays with Four-Primary-Color LCDs," SID Symposium Digest 37, Book 2, 1221–1224 (2006).
9S. Ueki, K. Nakamura, Y. Yoshida, T. Mori, K. Tomizawa, Y. Narutaki, and Y. Itoh, K. Okamoto, "Five-primary-color 60-in. LCD with novel wide color gamut and wide viewing angle," SID Symposium Digest 40, 927–930 (2009).
10K. Yoshiyama, M. Teragawa, A. Yoshida, K. Tomizawa, K. Nakamura, Y. Yoshida, Y. Yamamoto, and N. Ohta, "Power saving: A new advantage of multiprimary-color displays derived by numerical analysis," SID Symposium Digest 41, 416–419 (2010).
11M. Yamaguchi, H. Haneishi, and N. Ohyama, "Beyond red-green-blue (RGB): spectrum-based color imaging technology," J. Imag. Sci. Technol. 52, No. 1, 010201 (2008).
12Y. Murakami, J. Ishii, T. Obi, M. Yamaguchi, and N. Ohyama, "Color conversion method for multiprimary display for spectral color reproduction," J. Electronic Imaging 13, No. 4, 701–708 (2004).
13T. Ajito, K. Ohsawa, T. Obi, M. Yamaguchi, and N. Ohyama, "Color conversion method for multiprimary display using matrix switching,"Opt. Rev. 8, No. 3, 191–197 (2001).
14F. Konig, K. Ohsawa, M. Yamaguchi, N. Ohayama, and B. Hill, "A multiprimary display: optimized control values for displaying tristimulus values," Proc. PICS, 215–220 (2002).
15M. Derhak and M. Rosen, "Spectral Colorimetry using LabPQR: An Interim Connection Space," IS&T/SID 12th Color Imaging Conference, 246–250 (2004).
16R. L. Alfvin and M. D. Fairchild, "Observer variability in metameric color matches using color reproduction media," Color Res. Appl. 22, No. 3, 174–188 (1997).
17M. Yamaguchi, R. Iwama, H. Kanazawa, N. Fujikawa, H. Fukuda, H. Haneishi, N. Ohyama, H. Wada, T. Kambara, M. Aihara, Y. Yamakawa, A. Nemoto, M. Furukawa, and Z. Ikezawa, "Color reproducibility of skin lesions in multispectral video: Experimental evaluation," Proc. IS&T/ SID 14th Color Imaging Conference, 8–13 (2006).
18M. Yamaguchi, Y. Murakami, H. Hashizume, H. Haneishi, Y. Kanno, and Y. Komiya, "High-fidelity color video reproduction of open surgery by six-band camera," Proc. SPIE 7627, 762707 (2010).
19L. DeMarsh, "Colorimatry for HDTV," IEEE Trans. Consum. Electron. 37(1), 1–6 (1991).
20J. Kumada and T. Nishizawa, "Reproducible color gamut of television screens," SMPTE J. 101(8), 559–564 (1992).
21M. R. Pointer, "The gamut of real surface colors," Color. Res. and Appl. 5, No. 3, 145–155 (1980).
22M. Brennesholtz, "Expanded-Color-Gamut Displays – Part 2: Wide-Color-Gamut Displays," Information Display 22, No. 10 (Oct. 2006).
23J. Takiue, S. Sugino, Y. Murakami, T. Obi, M. Yamaguchi, and N. Ohyama, "Evaluation of smoothness in color gradation on multiprimary display," Proc. IS&T/SID 12th Color Imaging Conference, 257–264 (2004). •