Perceptual Science Series
Orientation-defined textures (ODT) are of a particular significance to low and intermediate level vision, and here we investigate them from a computational, psychophysical, and physiologicial perspectives. Geometrically, we show that two curvatures arise, which describes the textures local behavior and which are used to develop a theory of ODT "good continuation" that extends common terminology from Gestalt psychology. Computationally, we then use this theory for the contextual organization of globally (piecewise-) coherent flows from noisy and locally incoherent measurements, a biologically motivated computation which we apply and demonstrate on texture, shading, and color images. We further extend the formal analysis to novel psychophysical findings in orientation-based texture segmentation, as well as to a new functional explanation for basic aanatomical features of the primary visual cortex.