Stereo integration, mean field theory and psychophysics

Abstract

We describe a theoretical formulation for stereo in terms of the Bayesian approach to vision and relates it to psychophysical experiments. The formulation enables us to integrate the depth information from different types of matching primitives, or from different vision modules. We solve the correspondence problem using compatibility constraints between features and prior assumptions on the interpolated surfaces that result from the matching. We use techniques from statistical physics to show how our theory relates to previous work. Finally we show that, by a suitable choice of prior assumptions about surfaces, the theory is consistent with some psychophysical experiments which investigate the relative importance of different matching primitives.