Representing Colors as Three Numbers

GB in graphics is both a way of specifying color and a way of viewing color. Graphics algorithms manipulate RGB colors, and the images produced by graphics algorithms are encoded as RGB pixels and displayed on devices that render these pixels by emitting RGB light. Colored images are also used to specify color in graphics. These images may be captured by cameras or scanners, interactively drawn using tools such as Adobe PhotoShop, or algo-rithmically generated. But, what do all of these RGB values mean with respect to color perception? How does the RGB triple captured by a digital camera relate to the RGB pix-els displayed on a monitor? How does the RGB triple selected with an interactive color tool relate to the RGB triple used to color an object in a 3D rendering? Most computer graphics texts and tutorials provide a description of human color vision and measurement as defined by the CIE tristim-ulus values, XYZ. Often missing, however, is an in-depth discussion of the relationship between the different applications of RGB and XYZ, and any discussion of color models beyond trichromacy. The goal of this tutorial is to provide a complete, concise analysis of RGB color specification and its relationship to perceptual and physical specifications of color, and to introduce some models for color perception beyond tristimulus theory. That color can be represented by three numbers— whether RGB or XYZ—is a direct result of the physiology of human vision. Electromagnetic radiation whose wavelength is in the visible range (370 to 730 nanome-ters) is converted by photopigments in the retinal cones into three signals, which correspond to the response of the three types of cones. This response is a function of wavelength and is described by the spectral sensitivity curves for the cones, as Figure 1 shows. Colored light can be represented as a spectral distribution , which plots power as a function of wavelength. (Other fields, such as signal processing, plot spectra as a function of frequency, which is the inverse of wavelength .) The cones convert this to three cone response values (L, M, S)—that is, the cone sensitivities in the long, medium, and short wavelength regions—defined by integrating the product of the spectral sensitivity curves and the incoming spectrum. Figure 2 shows this process. Two important principles follow from this process: ■ Trichromacy: all spectra can be reduced to precisely three values without loss of information with …