Abstract

Spatial derivatives of moving and/or stereoscopic images constitute information about the differential structure of surfaces in 3-D. To investigate the role of such image properties, we have evaluated the visibility of spatial differential structure in both moving and stereoscopic images. The general method involves examining the effects of rapid random transformations of the 2-D images (translations, dilations, and 2-D and 3-D rotations) on detections and discriminations of 3-D surface structure. In stereoacuity experiments such transformations have been applied independently to the monocular half-images prior to binocular combination. Results of experiments on both structure from motion and stereoacuity indicate that (a) vision is very sensitive to the image deformations that carry information about surface shape, and (b) these deformation components can be visually distinguished from other differential components that have the same effects on the image motions and binocular disparities of both individual points and pairs of points. Thus, vision seems to be sensitive to the local differential structure of moving images—a four-parameter property, supported by relations among at least five neighboring image points.

© 1991 Optical Society of America