COMPUTATION AND PSYCHOPHYSICS OF MOTION AND STEREOPSIS

We would like to construct computational models of vision based solely of physiological mechanisms, with the long-term goal of understanding a large array of perceptual phenomena form a few underlying physiological principles. In the past few years, we have been investigating physiologically realistic models of stereovision and motion-stereo integration. We now propose to continue these studies along with several major new directions in order to achieve a deeper and more comprehensive understanding of binocular vision. We will first generalize our current one-dimensional stereo model for horizontal disparity computation to both spatial dimensions, and apply the model to explain the puzzling properties of depth perception from vertical disparity including the well-known induced effect, and the disparity attraction and repulsion phenomenon observed on two-dimensional stimuli. We will then propose a physiological algorithm for directly computing disparity gradients, and use it to account for the observed disparity-gradient limits for binocular fusion, and the paradoxical perception of disparity gradients in the absence of ordinary disparity cues. We will also analyze the relationship among different stereo algorithms, and examine how the psychophysically and physiologically motivated models are related to each other. In addition, we will construct three-dimensional, spatiotemporal receptive field profiles for binocular cells in the primary visual cortex based on the available physiological data, and use them to stimulate and explain the disparity tuning behavior of simple and complex cells in response to a variety of static, dynamic, and moving stimuli that are commonly used in physiological experiments. Finally, we will conduct psychophysical experiments to test some new predictions from our models, and to gain useful information for guiding our future modeling efforts. These studies will address several long-standing and challenging issues in vision research from a physiological perspective. Together with our previous work, they will provide a coherent understanding of many seemingly unrelated phenomena in binocular vision and motion perception.