Despite a long history of distinguished research, our understanding of how humans perceive the shapes, sizes, and locations of objects in the three-dimensional space surrounding them is still incomplete. In particular, we are only beginning to understand how motion of an image on the retina, resulting from movement of the object being observed or of the observer, leads to a perception of object shape and of spatial layout of a complex scene. One reason that this aspect of space perception is only now receiving due consideration is that proper investigation of the question was impossible until recently, when fairly powerful graphics computers became widely available. This research will address how adult human observers are able to perceive three-dimensional shape solely from retinal image motion as might be produced by a rapid succession of flat images on a computer video display. The primary goals are to identify which aspect of the retinal image motion is critical for the perception of 3-D shape and to develop a computational model that can predict the perceived shapes corresponding to different patterns of retinal motion. Another goal is to determine the degree to which the perception of 3-D shape, once built up, perseveres in the absence of sustaining retinal stimulation; there is evidence that some sort of internal representation does remain even when sensory stimulation is interrupted. A final goal is to determine how the perceived shape produced by retinal stimulation depends upon the perceived distance of the stimulus; prior work suggests that the perceived shape of a simulated object undergoing rotation varies in a different fashion with distance of the simulated object than the perceived shape of such an object undergoing side-to-side motion (translation). If confirmed, this result will provide important clues to the nature of the brain process involved in the perception of three-dimensional shape from motion. Enhanced understanding of these processes may lead to two important applications. First, it will indicate ways of optimizing displays for use in scientific visualization; scientific visualization refers to the use of 3-D animations as tools in understanding abstract scientific concepts (e.g., electron cloud), structures (e.g., complex molecules), and processes (e.g., molecular interactions). Second, enhanced understanding of how humans recover shape from motion will point to one possible way for computer visual systems to exploit optical motion in the analysis of complex scenes. In addition, understanding human perception of shape from motion will contribute to understanding of the more general problem of visual space perception, mentioned above.
