Abstract

The 3-D shape of a surface is often revealed by the shape of curves or lines that lie across the surface. The immediacy with which surface shape emerges from even simple line drawings demonstrates the existence of a perceptual process that effectively backprojects 2-D image contours into 3-D. Since the backprojection is not unique, the perceptual process must incorporate specific constraints stemming from certain assumptions made by the human visual system. A theory for these constraints^1,2 shows how both intrinsic geometry and surface orientation can be inferred, and suggests image configurations such as contour intersections and parallelism, that may trigger these inferences. To determine how these constraints are embodied in a perceptual process, it is necessary to first understand the output representation that is developed by that process. How is the surface shape information represented? Insight into this question is emerging by examining the interactions between stereopsis and the monocular processing of surface information.

© 1986 Optical Society of America