Shape Processing as Inherently Three-Dimensional

Abstract

A primary goal of visual encoding is to determine the inherently three-dimensional (3D) shape and motion structure of the visual scene to provide a functional representation of the scene layout and trajectories of the objects within it. A surface representation is a natural means of encoding the information from the sparse visual cues available, and psychophysical data suggest that surface interpolation is only possible through 3D interpretation of the depth structure provided by local cues; luminance-based interpolation fails in the absence of a 3D interpretation (whether from 2D or 3D cues). Functional imaging studies suggest that this 3D interpretation is located at the dorsal extreme of the lateral occipital complex. The neural 3D interpretation is limited by smoothness constraints derived from stereoscopic studies, although piecewise discontinuities are also permitted. The dimensions of its representational space, however, are not easily conceptualized in the 3D space that the shapes inhabit, but require the much larger configurational space of all recognizable 3D shapes. In this context, ‘shape’ is a conceptual abstraction of those surface configurations specifiable within our relatively limited cognitive window.KeywordsDepth ProcessingDepth StructurePictorial DepthLateral Occipital ComplexLuminance ProfileThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.