ABSTRACT
This chapter shows that a biologically plausible laminar cortical model that explains many psychophysiological phenomena can also be extended to do stereo matching on natural images without any prior learning. The model proposes how visual cortical areas are defined in terms of layered circuits and how these laminar circuits interact using bottom-up, horizontal, and top-down connections to generate 3D boundary and surface representations whose properties simulate those of conscious 3D surface percepts. The 3D LAMINART model builds upon the discovery that the visual cortical streams that process perceptual boundaries and surfaces obey computationally complementary laws. The 3D LAMINART model was developed to explain and predict perceptual and neurobiological data in terms of how laminar cortical mechanisms interact to create 3D boundary and surface representations. The 3D ARTSCAN model has extended the competence of 3D boundary and surface computations to an active vision framework wherein eye, or camera, movements freely scan a 3D scene while perceiving and learning to recognize it.
