Abstract
Humans generally perceive approximately constant object colors despite large variations in illumination. Color constancy is presumably mediated by several mechanisms operating on different levels, ranging from photoreceptor adaptation to contextual effects. We have recently shown that color appearance of familiar objects can be biased towards that object's typical color, which might be an additional mechanism for color constancy. Here we investigate whether these top-down influences are robust to changes in illumination. We asked 10 observers to adjust the color of eight fruit and vegetable objects displayed on a monitor to either their respective typical colors or to an achromatic color. In addition, observers made the achromatic settings with homogeneous discs and 1/f noise patches. We used a neutral illuminant (D65) and four other illuminants that were taken from the cardinal axes of DKL color space (bluish, yellowish, greenish and reddish). The achromatic settings generally shifted close to the illuminant chromaticity. Observers showed nearly perfect color constancy under the bluish, greenish and reddish illuminants, and good constancy under the yellowish illuminant. Constancy was best with the disc stimuli. The fruit achromatic settings shifted with the illuminant nearly to the same degree as the disc settings, and the fruit typical settings shifted as much as monitor gamut allowed. As previously found with a neutral illuminant, there was a bias in the fruit achromatic settings towards the opposite direction from the typical settings under all illuminants. Thus, observers had to compensate for the top-down influences of object identity on color appearance regardless of lighting condition. The robustness of the memory color effect indicates that it is a plausible additional mechanism for color constancy, acting by pushing color appearance of a familiar object towards its typical color under any illuminant chromaticity.
Supported by German Science Foundation Ge-879/5.