Spatial constraints of stereopsis in video displays

Abstract

Recent development in video technology, such as the liquid crystal displays and shutters, have made it feasible to incorporate stereoscopic depth into the 3-D representations on 2-D displays. However, depth has already been vividly portrayed in video displays without stereopsis using the classical artists' depth cues described by Helmholtz (1866) and the dynamic depth cues described in detail by Ittleson (1952). Successful static depth cues include overlap, size, linear perspective, texture gradients, and shading. Effective dynamic cues include looming (Regan and Beverly, 1979) and motion parallax (Rogers and Graham, 1982). Stereoscopic depth is superior to the monocular distance cues under certain circumstances. It is most useful at portraying depth intervals as small as 5 to 10 arc secs. For this reason it is extremely useful in user-video interactions such as telepresence. Objects can be manipulated in 3-D space, for example, while a person who controls the operations views a virtual image of the manipulated object on a remote 2-D video display. Stereopsis also provides structure and form information in camouflaged surfaces such as tree foliage. Motion parallax also reveals form; however, without other monocular cues such as overlap, motion parallax can yield an ambiguous perception. For example, a turning sphere, portrayed as solid by parallax can appear to rotate either leftward or rightward. However, only one direction of rotation is perceived when stereo-depth is included. If the scene is static, then stereopsis is the principal cue for revealing the camouflaged surface structure. Finally, dynamic stereopsis provides information about the direction of motion in depth (Regan and Beverly, 1979). Clearly there are many spatial constraints, including spatial frequency content, retinal eccentricity, exposure duration, target spacing, and disparity gradient, which - when properly adjusted - can greatly enhance stereodepth in video displays.