Shape from Shading, Occlusion and Texture

Abstract

Vision is a highly complex process. There is an important distinction between those visual processes which involve high level, or semantic, information and early vision processes which do not use such knowledge. A primary goal of an early vision system, be it human or mechanical, is to determine and represent the shape of objects from their image intensities. Marr (1982) calls such a representation, which makes explicit the distance to, and orientation of, the visible surfaces from the standpoint of the viewer, a 2–1/2D sketch. He describes several independent processes, or modules,which compute it. Marr’s research focussed on stereopsis and structure from occlusion boundaries and shape from texture.KeywordsPrincipal CurvatureSurface OrientationReflectance FunctionPhotometric StereoSpecular ComponentThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.