The retinal image of a textured three-dimensional surface contains variations that the visual system can use to extract the shape of the surface. We found that for the projected image to convey shape, it must contain correlated changes in oriented energy along lines corresponding to the projected lines of maximum curvature of the surface. We will develop a model that extracts these correlated changes in oriented energy using associative mechanisms in a space-spatial frequency analysis. These changes in oriented energy will be used to compute local surface normals to extract the global shape of the surface. In addition, we will: 1) corrugate and analyze the Brodatz collection (1966) to examine oriented energy in natural textures, 2) elaborate our previous technique of measuring ordinal depth along the surface to measuring local curvature to examine observers perceptions of surfaces of different shapes, 3) use isoluminant grating and plaid patterns of various frequencies to measure shape-from-texture for colored patterns.
