On the accuracy of surface reconstruction from disparity interpolation.

Observers viewed flashed random-dot stereograms depicting a pair of long, narrow, curved ribbons of textured surface defined by a Gabor function in disparity. Observers judged the location of the peak of the depth profile of one ribbon relative to that of the other. In one ribbon, disparity changed smoothly while in the other disparity was periodically sampled. Up to a limiting sampling period, disparity interpolation produced accurate surface reconstruction, but beyond that performance deteriorated rapidly. This interpolation limit depended on surface orientation (vertical vs horizontal) and disparity sign, but not Gabor spatial frequency.