Efficient Computations and Representations of Visible Surfaces.

Abstract : When the visual system is approach as performing information processing tasks, the processor can be characterized as solving a computational problem (Marr, 1976; Marr & Poggio, 1977). The first step in this approach is to formulate the problem precisely. This report addresses two such problems, the extraction of surface shape information from the image of contours lying on a surface and the determination of transparency. Our preliminary analyses discuss the constraints and assumptions that underly the perception of surface shape and transparency. The identification and elucidation of these assumptions and constraints is important for the design of flight simulators, in order that synthetic images are not built up from conflicting ones that will mislead the observer. The first part of the report examines our perception of surface contours (e.g., the projections of linear surface markings such as seams, wrinkles, pigmentation edges, glossy reflections, and shadow edges). Generally such contours lie interior to the silhouette of an object and have been regarded as merely contributing to texture, or to making the surface appear solid, or simply adding to the complexity of the image. In fact, they constitute an important source of information about surface shape.