Most theories of colour constancy assume a flat coloured surface and a single homogenous light source. Natural situations, however, are 3-dimensional (3-D), are hardly ever restricted to a single light source, and object illumination is never homogenous. Here, two special cases of secondary light sources with sharp boundaries were simulated on a computer screen: a house-like 3-D object with colour patches in sunlight and shadow, and a Mondrian-type pattern with a coloured transparency covering some of the colour patches. Subjects made `paper'-matches between colour patches in light and shadow and between patches under the transparency and without the transparency. Matching did not depend on whether the simulated lighting condition was natural (yellow light, blue shadow) or artificial (green light, magenta shadow). Patches under a coloured transparency produced lightness constancy but subjects could not discount chromaticity shifts induced by the transparency. The number of context patches (2 vs 6) made no difference, and it made no difference whether the transparency covered the Mondrian completely or only partially. These results indicate that subjects were not able to use local contrast cues at sharp illumination boundaries to discount for the illuminant.
Dieser Eintrag ist Teil der Universitätsbibliographie.