Natural Color Matrix: Delivering Color Fidelity

By David Lieberman

How important to a product is the image quality of its liquid-crystal display (LCD) screen? In the early days of flat-panel displays (FPDs), front of screen performance was not a major perceived value in most applications. In recent days, however, OEMs have come to realize that the screen is the face of their products and, indeed, the face of the company.

Increasingly, a more and more savvy user base is making its purchase decisions based on the quality of a product's screen. In automotive applications and elsewhere, manufacturers are learning that inferior displays result in lost sales, while superior displays provide true differentiation that customers have come to appreciate.

A new technology developed by Mitsubishi Electric, and now being incorporated into LCD screens from Optrex America Inc., provides unprecedented color fidelity in FPDs. Dubbed NCM for Natural Color Matrix, it consists of a patented color transformation algorithm, implemented in hardware, that provides real-time, on-the-fly processing to precisely match the colors specified in a data source.

Alternative software-based color management schemes, based on color look-up tables and simple matrix mappings, have proven themselves unable to deliver the color fidelity achievable by a true transformation calculation. The difference NCM brings to LCD applications is immediately apparent to the eye in the vivid color renditions provided by the technology, as well as crisp text and artifact-free motion video. A side-by-side look at two comparable LCD screens, one with NCM and one without it, dramatically demonstrates the visual effect of true color fidelity vs a compromise solution. (See Figure 1: The NCM Difference)

Figure 1: The difference NCM makes in color fidelity is readily apparent to the human eye.

The Color Challenge

Color has three components, known as HSI for hue, saturation and intensity; multi-color displays conventionally contain three primary color stripes or triads of red, green and blue (RGB). The conventional metric for measuring the difference between a specified color and a displayed color is Δ E*, with larger numbers meaning greater divergence between the two.

With the proliferation of e-commerce in the 1990s, it became clear that inconsistency in color rendition by display monitors was creating a problem. In Internet shopping, the major cause for merchandise returns was found to be that the color of the item purchased was different from the color of the item displayed on a screen. To address this issue, an international standard (IEC 61966-2-1) was created (and adopted in 1999) for color management in multimedia systems, defining a default color space known as sRGB.

So far, so good ...for color matching in an era when the cathode-ray tube (CRT) was the dominant display technology for monitors. But then came the explosion in desktop LCD monitors, as well as in a wide variety of commercial and industrial applications for LCDs, creating new problems in color matching.

One problem is that the sRGB specification is geared to CRTs, but FPDs have completely different color characteristics. The result is readily apparent out in the real world today: gray hamburgers and pink ketchup on the drive-through LCD screens outside fast food restaurants, for example; lost differentiation of tissue type on diagnostic medical monitors; and miscolored sweaters (perhaps soon to be returned) on the displays of e-catalog kiosk systems.

Color imaging on FPDs has, indeed, suffered from display differences and a one-size-fits-all sRGB standard. A CRT is an emissive display, its light created by electrons striking red, blue and green (RGB) phosphors; while LCDs are transmissive displays, operating by blocking the light from a backlight or letting it pass through color filters. The color point of white light from a CRT monitor (specified in K for degrees Kelvin) is around 9000 to 10,000K, while the figure for an LCD is typically 7000 to 8000K. The white point for television sets is 6800K.

The first step in reproducing the input colors from a data source as output on an FPD requires massaging the data to suit the native white point of the display in question. For an LCD, the simplest solution is merely to rotate the display's color space, defined by the CIE (Commission Internationale de l'Eclairage) chromaticity chart, to suit the white color point of the backlight. (Figure 2. Color Conversion Differences) This technique, though, automatically limits dynamic range.

Figure 2: Conventional techniques (a) sacrifice part of the color potential of an LCD, which NCM (b) recaptures.

With this common chromaticity-shifting technique, much of the original blue data is eliminated from the FPD's color pallet. In the implementation illustrated in Fig. 2, for example, the 8 bits of blue in a 24-bit file would shrink to about 3 bits. This truncation of the original data for one color reduces the number of linear gray-scale steps for all colors between full on (white) and full off (black). Thus, a data file built using a pallet with 256 levels of gray (and millions of colors) will have only 32 or 64 gray levels and a few thousand colors to work with. The likelihood that an input color won't find a matching output color is very high.

Flawed Solutions

Conventional color management techniques create a color look-up table that maps the colors in a data source to the particular characteristics of a display, whether it's an LCD, for example, a plasma display panel (PDP) or an organic light-emitting diode (OLED) display. A full 24-bit color pallet (8 bits each for the red, green and blue primaries) contains over 16 million colors and, thus, a look-up table providing all the color options would be too memory intensive and cost prohibitive. Therefore, compromises are frequently made, such as settling for an 18-bit color look-up table, which limits the pallet to just 262,144 colors.

In this scenario, however, there are techniques for coping with a 24-bit data source. A simple matrix calculation, for example, can be applied to the look-up table data to approximate the missing 6 bits and attempt to match the colors specified in the source. The application of this technique, however, has been severely flawed.

Simple matrix calculations handle HSI as a single entity, without separating the chromic (H) and achromatic (SI) aspects of data. The problem here is that changing the achromatic characteristics of one color component changes those characteristics of the other components, which changes the resultant RGB color, and color management becomes very complex and unpredictable. Simply put, without independent chromatic/achromatic control, lowering the brightness of one color on a transmissive LCD increases the relative brightness of the other two colors.

Figure 3: NCM separates chromatic from achromatic data, giving
independent control over individual colors.

NCM, in contrast, provides independent control over the chromatic and achromatic characteristics of a color construct, applying a transformation operation that remains true to the original source. The sRGB allows a Δ E* deviation from color source to color image of 10 typical and 15 maximum. NCM achieve values of 0.38 typical and 0.65 maximum--a dramatic and easily discernable difference.

The NCM technology has already proven itself out in the field, implemented in electronic projectors being mass produced by NEC-Mitsubishi. In direct-view LCDs from Optrex, it's being matched with a 6500K backlight that provides an optimal balance between blue for text quality and red for accurately reproducing images.

A True Transformation

Requiring neither massive horsepower, large amounts of circuitry or burdensome memory resources, the Mitsubishi NCM algorithm is implemented in a small portion of a display module's timing controller (TCON) and it requires a small amount of ROM. At the factory, an NCM LCD is characterized, and the resulting data is used to create a unique set of constants for that display, based on its particular color filter characteristics, for example, the nature of its backlight and other factors such as whether or not it contains an anti-reflection filter.

These display-specific values are stored in mask ROM and used by the algorithm for color mapping. Unlike conventional color mapping processes, in which adjusting one color influences the other colors, NCM provides independent control. The algorithm operates on an incoming data stream on the fly, in real time, on a pixel-by-pixel basis. It is as capable of displaying full-motion video as it is of handling text and still images, while techniques based on look-up tables can easily get bogged down at video rates and one-size-fits-all algorithms fail to render colors accurately.

Figure 4: LCDs with NCM color conversion accurately reproduce colors as specified.

In operation, the NCM algorithm first separates input image data into chromatic data and achromatic data. Based on the stored values for the display, it then converts the chromatic data into a "hue region" consisting of six component colors (RGB, plus cyan, magenta and yellow) and an "inter-hue region" defining six more. Finally, using the independent achromatic data, it manipulates the chromatic data to generate the appropriate output color on the display. Changing the saturation and intensity of one color has no effect on the others. (See Figure 3: NCM Color Conversion)

The Application Space

There are a number of application areas for FPDs today in which color fidelity is important and a few where it's absolutely critical. Yet, across many arenas, digital content is being created based on the standard sRGB color space, and the content as displayed on FPDs does not match. An arbitrarily rotated color space produces unpredictable (and sometimes bizarre) color consequences, while a true transformation accurately reproduces the intent of the original content creator.

In the fast food arena, where the drive to increase revenue per customer is strong, images of gray milkshakes are a disincentive to buy.
In other retail environments, washed-out colors on a display do not help promote clothing sales or custom satisfaction when the real colors do not match.
In gaming applications, vivid colors produce a much more dramatic effect, as intended by the original content creator.
In test and measurement equipment, accurate color rendition is a genuine boon to the quick absorption of data by the human eye and brain--a critical feature for process control systems, where red warning messages should not flash pink.
And in diagnostic and imaging equipment, if some variable is represented by bright orange, it'd better not be displayed as a dull brown; nor should the color-coded alarm messages on patient care monitors diverge from caregivers' color expectations.

More and more, OEMs are discovering that better displays promote products sales and superior color rendition is a major component of imaging quality.