Computation and Psychophysics of Motion and Stereopsis

We would like to construct computational models of visual perception based on physiological mechanisms, with the long-term objective of understanding a diverse array of psychophysical and physiological data from a few unifying principles. In the past several years, we have created physiologically-based models to account for depth perception from horizontal disparity, vertical disparity, interocular time delay, and motion fields, and have demonstrated the explanatory and predictive power of our approach. We now propose to :ontinue these studies along several major new directions. We will first evaluate some recently proposed modifications to the disparity energy model, and suggest alternatives that will provide the most compact, complete and plausible summary of V1 disparity data to date. We will then examine the functional implications of the modifications, particularly the role of the proposed monocular nonlinearity in the computation of stimulus disparity maps. In addition, we will introduce neuronal noise and a decision rule into our most recent stereo algorithm to simulate disparity increment threshold (stereoacuity). We will apply this procedure to explain comprehensive stereoacuity data for simple gratings, compound gratings, and random- dot stereograms, and to account for the scale interactions in the data. Furthermore, we will extend our stereo model to explain depth effects of monocular occlusions (a phenomenon termed da Vinci stereopsis), and investigate how the conventional stereopsis and da Vinci stereopsis may be understood in a unified framework. Finally, we will model interactions between stereovision and vergence eye movement in a new framework of visuomotor integration, by combining the theories of Kalman filtering and stochastic optimal control with realistic disparity representation. We will also examine if our visuomotor model can provide a unified account for the delay-compensation responses, the memory activities, and the predictive receptive field shifts observed in parietal and frontal neurons, and for the psychophysical observation of enhanced stereoacuity by vergence and increased visual sensitivity at the next saccadic goal. These studies will address several long-standing and challenging issues in-vision and visuomotor integration from new, physiological perspectives. They will provide deeper insights into the fundamental mechanisms of visual processing that in the long term, may help build neuromorphic systems to aid people with impaired vision.