Lightness Perception Can Be Affected by Surface Curvature from Stereopsis

Abstract

It is demonstrated that lightness perception can be affected by shape from stereopsis. The starting point was a report by Knill and Kersten that the perceived lightness of a monocularly viewed surface can be affected by outline-contour cues indicating that the surface is three-dimensional (3-D). In that study stimuli consisted of two equally sized abutting regions each having the same vertical linear-intensity ramp, so that the horizontal abutting boundary of the two patches created a sharp change in intensity. When this version of the Craik-O'Brien - Cornsweet stimulus has a rectangular outline, it exhibits the standard simultaneous contrast illusion: equivalent patches in the top and bottom regions appear to have different brightness despite having the same luminance. Knill and Kersten replicated this phenomenon with stimuli whose outline-contour cues were consistent with a flat (planar) surface. They found, however, that the illusion was greatly reduced in stimuli with outlines consistent with two abutting 3-D quarter cylinders, for which equivalent regions in the two halves appeared of similar lightness. Knill and Kersten interpreted this effect in terms of surface-lightness computations that took into account 3-D surface shape to achieve an integrated interpretation of the luminance and shape data. In the present report three experiments are described for which these earlier findings were taken as the starting point. In the first experiment the results were replicated by the use of a different methodology. In the second experiment it was shown that shape-from-stereo can produce similar effects on lightness perception to that caused by shape-from-contour. Real 3-D objects with curved surfaces, luminance profiles of the Knill and Kersten type, and carefully controlled outline-contour cues were used so that the objects appeared flat when viewed monocularly but curved in 3-D when seen binocularly. The third experiment was a control confirming that the stereo effect was not simply due to differences caused by monocular versus binocular viewing. It is concluded that the human visual system uses stereo cues, as well as outline-contour cues, in the interpretation of luminance data to recover surface lightness.