Abstract
INTRODUCTION: Although reconstructing shape from shading is an ill-posed problem which has multiple solutions, human visual system selects a shape from a shading easily. For example, a disk with shading definable either a spherical or saddle shape, is mostly perceived as a spherical shape. This selection is often inconsistent with the assumptions for perceiving shape from shading, such as light-from-above assumption. What determines the selection, experience, physical validity or other factors? In this study, we investigated the effect of experience on the selection for perceiving shape from shading.
EXPERIMENT: An experimental session consisted of three blocks; two test blocks and a learning block between the test blocks. In the test block, observers viewed monocularly the shading stimuli simulating a saddle shape, which were perceived as spherical mostly. In the learning block, they viewed the same shading stimuli binocularly with disparity indicating the same saddle shape. Thus observers experienced the specific relation between the shading and the saddle shape from disparity in the learning block. They responded perceived shape in all blocks. We compared the perceived shapes in pre- and post-learning test blocks to show whether there was the effect of the learning for selecting shape from shading. The session was repeated five times in different days. Seven observers participated in this experiment.
RESULT: Three subjects became to perceive the saddle shape in the test block after the learning block. This result indicates that they learned to perceive the saddle shape for the stimuli with shading only. However the other subjects didn't. Two of them couldn't perceive the saddle shape in the learning block, so it is reasonable that they didn't learn to perceive the saddle shape. These results suggest that the selection of perceived shape from shading is, at least partly, empirically determined.