Abstract
In the analysis of shape from surface contours, researchers have proposed a variety of assumed constraints by which contour patterns are perceptually interpreted. It can be assumed, for example, that the contours are planar cuts through a surface (Tse, 1999), or that they lie along surface geodesics or lines of principal curvature (Stevens, 1981; Li & Zaidi, 2000). This paper will present a number of clear counter-examples for each of these proposed constraints, and it will consider the effects of several different stimulus factors that can influence observers' judgments of 3D shape from surface contours. These factors include the manner in which a surface is textured — i.e., mapped textures, such a wall paper, behave quite differently than textured surfaces that have been sculpted out of a solid material, such as wood or marble. Other factors that can influence performance include the sign and magnitude of surface curvature, how it is oriented with respect to the surface texture, and its position and orientation in space relative to the point of observation. Our results reveal that some of the stimuli used in previous investigations have been highly non-generic, and may therefore have provided a misleading picture about how surface texture is perceptually analyzed. The evidence suggests that perception of shape from surface contours is surprisingly robust. Observers' perceptions of 3D shape are “qualitatively” correct over a wide variety of contour generation procedures, provided that the projected images of those contours are curved.