Chapter Three: Perceiving and Representing Shape and Depth

To determine how to represent the shape and depth of a layered transparent surface most effectively with sparse, opaque texture, we need both to understand the various potential sources of shape and depth information available in a two-dimensional representation of a threedimensional scene and to have some sense of how effectively our visual system might use this information to accurately infer the shape and depth relations of the depicted objects. The distinction between shape cues and depth cues has historically been somewhat blurred, and it has not been until very recently that a coherent theory for the definition of object shape in a global sense, distinct from local measurements of slant and tilt, has emerged [Koenderink 1990], although the question of how our perception of the global shape of a surface might be related to our perception of the depths and local orientations of that surface at isolated points is still not exactly clear [Bülthoff and Mallot 1988]. While there is insufficient time and space here to do justice to the vast extent of current knowledge about shape and depth perception, I have attempted in this chapter to review selected aspects of these topics that I hope will be useful for providing insights into methods for more effectively representing shape and depth information in computer-generated images.