What happens with our color constancy performance when varying the chromatic distribution
of a stimulus while keeping average chromaticity fixed? We synthesized images composed of
about 900 color patches varying in CIE L*a*b* values. Both the average chromaticity and the
chromatic distribution were varied. The 2D Gaussian distribution of a*b* values was either
circular, or ellipsoid with the variance in a* 5 times that in b*, or vice versa. Four
illuminants (equidistant from the neutral point) were used to simulate illumination of the
color patches by daylight variants. Using triad-comparison, observers judged the color
naturalness of the scenes under two illuminants (6 illuminant pairs possible) against the
scenes under neutral reference illumination. They indicated which of the two scenes
(illuminants) reproduced the colors of the reference scene most closely. Our results show
that when the dominant axis in the chromatic distribution is parallel to the direction of
the illuminant change, color constancy is best. We conclude that color constancy depends on
the chromatic distribution of the scene.
@Article{LucassenPerception2008,
author = "Lucassen, M. and Gijsenij, A. and Gevers, T.",
title = "Chromatic Distribution Affects Color Constancy",
journal = "Perception",
number = "supplement",
volume = "37",
pages = "147",
year = "2008",
url = "https://ivi.fnwi.uva.nl/isis/publications/2008/LucassenPerception2008",
has_image = 1
}