Color Appearance Models

CIELAB was a huge step: a perceptual space where distance roughly tracks perceived difference. But it makes a simplifying assumption - that a color is judged only relative to a single reference white, in isolation. Real viewing is messier. The absolute light level matters (the same paint is more vivid outdoors than in a dim room). The surround matters (an image looks lower-contrast in a dark cinema). The background a color sits on matters. CIELAB sees none of this; it returns the same L*a*b* regardless.

A color appearance model (CAM) closes that gap. You give it the color and the viewing conditions, and it predicts perceptual appearance correlates - how light, how colorful, what hue the color will actually look. CIECAM02 and its successor CIECAM16 are the standard models. Think of a CAM as "CIELAB that knows what room you are in."

A CAM does not output one set of coordinates - it outputs a richer family of perceptual attributes, and the distinction between the relative and absolute ones is the key insight.

What makes a CAM different from CIELAB is that the viewing condition is a first-class input. To run CIECAM you specify the parameters of the scene as well as the color:

Feed those in and CIECAM applies a chromatic adaptation transform, a cone-response nonlinearity, and a set of equations that produce the correlates. The inverse model is just as important: given a target appearance and a different set of conditions, it computes the stimulus that reproduces that appearance - the basis of cross-media color matching.

Two effects of light level are why a CAM needs absolute luminance. The Hunt effect: colorfulness increases with luminance - colors look more vivid in bright light, washed out in dim light. The Stevens effect: perceived contrast increases with luminance - the difference between light and dark stretches as it gets brighter, and compresses in the dark. Raise the light level and watch both at once.

The surround - what is around the display, not in it - changes perceived contrast too (the Bartleson-Breneman effect). The same image looks lower contrast in a dark surround (a cinema) than in a bright one. This is why content graded for a dark theater is encoded with extra contrast (a higher system gamma): the dark surround will visually flatten it back to normal. Switch the surround to see the compensation a CAM prescribes.

The first stage of any CAM is a chromatic adaptation transform (CAT) - the math version of the white-balance your eye does automatically. It scales the cone responses so the adapting white maps to "white," letting the model predict corresponding colors: the stimulus under illuminant B that looks the same as a given color under illuminant A. This is exactly what converting an image from a D50 print viewing condition to a D65 screen needs.

Six questions on appearance correlates, viewing conditions, and the named effects. Instant feedback, no scores recorded - a wrong answer comes with a short explanation.

Appearance models extend the measured spaces with viewing conditions. The numbers reflect each article's position in the editorial roadmap.