Expanded-Color-Gamut Displays – Part 2: Wide-Color-Gamut Displays
Newer display technologies, including microdisplay-projection and LCD TV, are not limited by the color-gamut restrictions of phosphor-based displays, such as CRTs, and can have expanded color gamuts covering most natural colors. Evaluation and comparison of these expanded color gamuts is difficult, however, and no standard method is used by the display industry. In Part 2 of this two-part series, several proposals and recommendations are made to standardize the way display color gamuts are compared to each other.
by Matthew S. Brennesholtz
HIGH-DEFINITION (HD) TV sets with color gamuts larger than the specified HDTV video color gamut are becoming common in the consumer-electronics industry. Part 1 of this series, which appeared in the September issue of Information Display, presented the basics of color science and an introduction to the color encoding of video signals. This month, we will examine the colorimetry of the displays themselves and methods to compare different display color gamuts.
Real Color Gamuts
One question to be answered before looking at normal- and wide-color-gamut displays is, How wide must a display color gamut be to give viewers a vivid and realistic view of the world? The answer to this question is not simple. At its core, the answer is related to artistic and marketing factors, not primarily engineering and scientific ones.
Figure 1 shows several real color gamuts, listed in order of increasing size. These gamuts are described in more detail in Table 1 which gives the total area of each gamut in both x-y and u′-v′ space, plus the fraction in each system of the total color gamut produced by pure spectral colors.
The two surface color gamuts represent a line drawn in x-y space that encircles all colors measured on a large collection of real objects by Pointer.1 His real objects included not only natural objects such as butterflies and flowers, but man-made objects such as paint samples, fabrics, and the Munsell color sample chips.
Matthew S. Brennesholtz is Senior Analyst at Insight Media, 3 Morgan Ave., Norwalk, CT 06851; telephone 203/831-8464, e-mail: matthew@insightmedia.info.
Fig. 1: CIE 1931 color diagram with real color gamuts.
The L* effect that differentiates the Surface-1 and Surface-2 gamuts comes from the L*u*v* or L*a*b* color spaces which are based on, but are more complex than, the x-y or u′-v′ color spaces. These spaces are designed to take the relative luminance (L* ) into account when comparing colors. For colors that have very low luminances, such as very dark reds, greens, and blues, the color will appear less saturated to the human eye than its location on the x-y or u′-v′ color chart would indicate. For example, a very dark blue does not look blue at all to the human eye; it looks black or near-black. In the Surface-1 gamut, this factor is taken into effect. Typically, real objects that have very high saturation2 also have very low reflectivities. All the objects measured by Pointer that were between the Surface-1 and Surface-2 gamuts had such low reflectivity that when L* was taken into account, they appeared to be inside the Surface-1 gamut.
Surface-1 gamut represents the colors a person would see when looking around in a world illuminated by sunlight. As such, it would seem reasonable that these colors should be reproducible in a video system in order to give reasonably natural colors. Unfortunately, the HDTV color gamut was chosen based on legacy TV broadcast standards and cathode-ray-tube (CRT) phosphors and does not completely cover even this minimally acceptable color gamut.
The Surface-2 gamut ignores the L* effect and encircles all surface colors that Pointer measured, regardless of their brightness. Some of the additional colors might be perceivable to the human eye under certain circumstances; for example, if a spotlight were shined on a highly saturated but dark object with otherwise relatively dim surroundings.
The film color gamut is the gamut that can be produced by cinema film. As plotted in Fig. 1, the L* factor is ignored. This gamut also represents the gamut of non-film transmissive objects such as color filters. One of the most common transmissive objects to the average citizen is the red, yellow, and green traffic signal.
The outermost gamut, labeled Spectral Colors, is just the normal border of the CIE 1931 color gamut and is produced by monochromatic light sources such as lasers. The L* rule of thumb that saturated colors are dim does not apply to emissive colors such as lasers, light-emitting diodes (LEDs), and CRT phosphors. Emissive objects can be both very saturated and very bright simultaneously.
Normal-Color-Gamut Displays
All television color gamuts, including the HDTV color gamut, are based on the phosphors available for direct-view CRTs in the '70s, '80s, and '90s.3 When the U.S. HDTV color gamut was established in the '90s, I was disappointed that a legacy color gamut was chosen for color encoding. Even at the time, it was well known that this gamut could not reproduce colors that the consumer wanted to see. It was also well known that expanded-color-gamut displays with non-phosphor technologies would soon be competing with phosphor-based CRTs and plasma displays.
Comparing the BT.709-2 HDTV gamut in Fig. 2 with the Surface-1 gamut reveals sig-nificant areas where HDTV cannot reproduce the colors that exist in real life. The areas most visible are the cyan and yellow/gold regions as indicated in Fig. 2. The magenta region is also reproduced poorly, but this region is not as important in terms of generating a pleasing picture for the viewer. Even the relatively large NTSC gamut does not completely cover all gold and cyan regions that viewers want to see.
Wide-Color-Gamut Displays
Wide-color-gamut displays are displays that have larger color gamuts than the HDTV BT.709-2 color gamuts. Projection systems with these larger color gamuts have come on the market, and direct-view LCD TVs have been demonstrated at the SID Symposium and can be expected to hit the market soon.
The color gamuts for three expanded-gamut projectors are shown in Fig. 3. The largest gamut is the color gamut based on laser illumination. The color points in this figure are based on color calculations using the announced Novalux wavelengths. The next smaller gamut is a color-sequential Multiple Primary Color (MPC) LCoS projector with five primary colors and an ultra-high-pressure (UHP) lamp. A typical three-primary-color digital-light-processing (DLP) system with a UHP lamp is also shown because it significantly expands the system color gamut compared to BT.709-2 in the green region.4
While the DLP projector has a relatively good green and an acceptable blue, it has a poor red point. This is typical of three-color systems with a UHP-type lamp: they can make good blue and green primaries, but if a good red is produced, the system brightness suffers dramatically. This is one of the advantages of a MPC system – a good red can be generated with a UHP-type lamp and make up for the low brightness by introducing a gold primary.5
Table 1: Real color gamuts
Name
One thing that should be noted in Fig. 3 is that wide-color-gamut projectors can produce colors that do not need to be displayed because they are outside the color gamut of real objects or even film. This is another advantage of MPC systems – the five (or more) primary colors can be tailored to produce all needed colors but few unnecessary colors.
"Unnecessary" colors are all relative. One must remember that video is an artistic medium with a strong commercial basis that caters to consumer tastes. As such, video does not always try to produce "real" colors because consumers have shown a strong preference for slightly exaggerated colors. They want pretty girls with red lips, plus green grass, and blue sky. Video also typically contains some entirely artificial colors produced by computer graphics.
Evaluation of Color Gamuts Using the u′-v′ System
As discussed in Part 1 of this series, the u′-v′ system, when compared to the x-y system, gives roughly equal weight to red and green, but undervalues blue. This is a major improvement over the x-y system and therefore should be used for color-gamut evaluation. One reason undervaluing blue makes little difference is in normal, three-color gamuts most of the gamut expansion in displays is done in the red and green regions. There is little freedom in three-primary systems for the display designer to expand the gamut in the blue. Since little change is expected in the color gamut in the blue, the accuracy of the evaluation in the blue is not too important.
In the past few years, display manufacturers have been measuring total color-gamut size in square x-y units or u′-v′ units and comparing it to the NTSC, BT.709-2, or some other defined color gamut. The problems of the x-y system have already been discussed, and the solution is clear: Do not use the x-y system to evaluate color-gamut size.6
When the total area of a display color gamut is evaluated in u′-v′ space, it makes little difference what reference is used – NTSC, BT.709.2, or some other – since these are just linear scale factors. Because the NTSC color gamut is currently used the most, I recommend it be used universally to simplify comparisons between numbers from different vendors, if total gamut area is evaluated.
The key question, though, is this: Is total gamut area the best way to evaluate color gamuts and compare them to each other? In this total-area system, unnecessary colors such as saturated magentas or greens with very high y values are given equal weights with cyan and gold colors that are critical to making good images.
Fig. 2: Video color gamuts compared to real colors.
Fig. 3: Expanded-color-gamut systems.
This measure would run from 0% for a monochrome display to 100% for a display that could accurately reproduce all the colors in the real color gamut chosen as a reference standard. Color gamuts with sizes larger than 100% would be impossible under this system.
In principle, this proposal is somewhat like the Color-Rendering Index (CRI) used for light sources. In the CRI system, certain standard reflective surface colors are viewed under sunlight and under the lamp to be evaluated. If all the standard colors are the same to the human eye under both sources, the CRI is 100%. The more colors that are visibly different under the test lamp and sunlight, the lower the CRI.
The first three columns of Table 2 show calculated color-gamut numbers of the MPC, DLP, and laser projectors plus Rec. 709 as a stand in for a "typical" CRT when Surface-1, Surface-2, and film are used as target color gamuts. The last two columns show the gamut area referenced to NTSC, without comparison to a reference real color gamut. The last column, where this comparison is done in x-y space, would be the number reported most commonly in the past.
One interesting thing to note is that if Surface-1 is used as a reference, all the color gamuts rate in the mid-to-high 90s. Even the BT.709-2 gamut, with its well-known deficiencies, rates at 92.2%. The conclusion is that all these systems can reproduce surface colors fairly well.
Another thing to note is the DLP projector rates fairly high in the traditional measure, comparison in x-y area to the NTSC gamut. When rated with this new system, however, it rates very close to the BT.709-2 CRT gamut. This occurs because the DLP set has very good greens but poor reds, and the x-y evaluation overweights the greens very strongly.
If we examine the film gamut, however, we see that BT.709-2 fares less well and can only reproduce 84.1% of the colors needed to give a true cinema experience to the viewer, while the laser projector can reproduce 96.7% of the colors needed. The colors the laser system cannot produce are mostly in the blue/cyan region. This is not surprising: it is virtually impossible for a three-primary-color additive color projector to reproduce the entire film gamut in the blue region and still maintain good yellows and greens. Laser projectors can also reproduce some colors not needed for the film experience, but that is irrelevant to this proposed gamut-evaluation system.
Recommendations for Color-Gamut Evaluation
The following recommendations are made for color-gamut-size evaluation:
1. Always evaluate in the u′-v′ color system, never in the x-y system.
2. Calculate the fraction of colors in a real color gamut that the display can reproduce. The preferred color gamut for this reference is the color gamut of cinema film.
There is a significant amount of work to be done on this modest color-gamut evaluation proposal. First, there is no standard "film" gamut. There are many types of film available, and they can be pushed to a slightly larger color gamut by pushing them to higher densities. The film gamut used in this paper is for a density of 3.0, while 3.5 is sometimes used to evaluate film. At a density of 3.5, however, the saturated colors in the film would be very dark if shown in a theater, and they would be dim enough so the L* effect probably would make them look less saturated than they actually are.
Second, this method needs to be standardized by a recognized standards body. There were many methods of measuring luminance, some giving wildly disparate values, before the ANSI luminance standard was written. Today, all projector makers talk about ANSI lumens in their technical papers at SID, and, hopefully, in all their marketing documents.
Finally, perception and human-vision experts should review alternatives to the u′-v′ system for color-gamut evaluation. This system undervalues blue. While this will make little difference in three-color systems, MPC systems are coming that can expand the color gamut in blue as well as red and green.
Since SID is not a standards-setting body, suggestions on how to generate a standard from this proposal are welcome.
References
1Surface-1 and Surface-2 color gamuts come from data given by M. R. Pointer, "The Gamut of Real Surface Colors," Color Research and Application 5:3, 145-155 (Fall, 1980).
2A color is called a saturated color when it is far from the white point and close to the outer border of the x-y or u′-v′ diagram.
3For those interested in how the relatively large 1953 NTSC color gamut evolved into the unacceptably small HDTV color gamut, see LeRoy DeMarsh, "TV Displays Phosphors/Primaries – Some History," SMPTE Journal, 1095-1099 (December, 1993).
4These last two color gamuts were measured by the author and reported in paper 64.3 in the 2005 SID Symposium Digest of Technical Papers.
5The introduction of new drive schemes and technology for ultra-high-pressure mercury lamps will reduce this problem somewhat in the future. See, Holger Mönch, paper 55.3, SID Symposium Digest of Technical Papers 37 (2006).
6See Seyno Sluyterman's Short Subjects column appearing in the May 2006 issue of Information Display. •
Table 2: Comparison of display color gamuts with various test methods