Three-dimensional perceptual grouping and filling-in of slanted and curved surfaces: Laminar cortical model and gloss perception experiments.

摘要

The first part of the thesis proposes a model for three-dimensional (3D) curved surface representation and 3D interpretation of two-dimensional (2D) images. Binocular disparity is a key cue for 3D perception and has usually been used to represent 3D planar surfaces in earlier models. It can also be used to glean information about slant or tilt of curved surfaces. Monocular cues also provide information about 3D structure of an object. Some monocular cues by themselves are ambiguous, but a combination of them can disambiguate the 3D interpretation. A neural model of how the visual cortex combines information from binocular and monocular cues to perceive and represent 3D curved surfaces is developed. The model includes cells that are sensitive to angles and disparity gradients and shows how such cells can be organized within the laminar circuits of cortical areas V1 and V2. The model is consistent with neurophysiological and psychophysical data. Model simulations show how horizontal connections can develop cell selectivity to angles, how 3D slanted surfaces are represented by cells that are sensitive to disparity gradients, how filling-in occurs across slanted surfaces, how a Necker cube 2D image can be represented in 3D, and how bistable Necker cube percepts occur.; The second part of the thesis describes psychophysical experiments on gloss perception. The perception of gloss is usually attributed to the presence of highlights on the object, which are dependent on surface properties, light source direction, and eye position. Even though a highlight is localized on an object, the perception of gloss is usually attributed to the whole object. How does a local highlight induce the perception of gloss at places far from the highlight? Does the perception of gloss at a particular point on the object vary as you move away from the highlight? The experimental results show that gloss perception is non-uniform on an object, varies with distance from the highlight, depends on surface distance rather than image distance, and is affected by discontinuities on the surface. An explanation of the results in terms of the FACADE theory of 3D vision is given.