Color inconstancy in CIELAB: A red herring?

Abstract

AbstractColor constancy is something we take for granted. An apple appears red despite monumental changes in the intensity and spectral balance of illumination. There are many components to the human visual system that make this illusion possible. Nonetheless, the system is imperfect. Objects will change apparent color under different illumination. The CIELAB color space allows for the computation of color values under any illuminant. CIELAB values will, in general, change as the illuminant changes. It would be natural to assume that these changes in CIELAB values would agree with the apparent color change that a human would experience. But the chromatic adaptation built into CIELAB is known to be less than optimal due to an imperfect emulation of how the adaptation process is performed in the human visual system. As a result, at least part of the illuminant color change implied by the CIELAB values is an artifact of this imperfect emulation. This paper investigates the magnitude of this artifact by comparing color changes implied by CIELAB with color changes in a color system that mimics CIELAB, except that it more closely emulates the human visual system. The difference between the two estimates of color change is assumed to be an artifact of CIELAB's poor emulation of the chromatic adaptation process. As color manufacturers are dealing with a wider spectrum of illumination sources, they may wish to consider the issue of color constancy; what colors and what formulations of colors exhibit better color constancy? Can CIELAB be used to quantify this? Or are color spaces that are based on LMS rather than XYZ more appropriate?