Abstract
We have constructed a model of lightness perception that contains just five free parameters and comprises two components: one considers a surface's luminance and the other the context in which the surface is placed. The luminance component assumes that the surface with the highest luminance is white and estimates the lightnesses of all other surfaces in the visual scene by comparing their luminances to the highest luminance in the scene. The context component of the model modifies these lightness estimates by taking into account the contrast between a surface and its neighbors. Depending on the exact luminance relationships, the context component will sometimes predict the lightness difference between a surface and its neighbors to be exaggerated (i.e. the simultaneous contrast effect) and sometimes be understated (i.e. an assimilation effect). In this way the context component of the model is able to modify the lightness estimates of the luminance component of the model so that they are more in keeping with human perception. The model is able to explain a number of visual phenomena including the simultaneous contrast illusion, illumination dependent errors, the Gelb effect, White's effect, the Benary cross, the White-Todorovic' illusion, the Koffka ring, the Craik-O'Brian-Cornsweet effect, a display containing multiple luminance steps as well as several important variations of the above displays.
This work was supported by a Helen Hay Whitney Foundation grant to P.H. and NIH grant EY 13135 and ARO grant 46961 to M.L.