Seminar Series

Binocular disparity in natural images

Binocular disparity provides important information about the three-dimensional structure of the environment. The geometry of binocular vision is well understood. In order to fully understand human binocular vision, it is important to complement geometrical analyses by considering the binocular disparities that will be expected in images of the natural environment. A simple environmental model will be presented that predicts the probability density functions and spatial distributions of horizontal and vertical disparities. The results of this model show that the distribution of disparities depends critically on fixation, and varies greatly form one image location to another. The results of this model will be considered in relation to Bayesian models of visual perception, and compared to well-established peculiarities of human binocular vision. Overall, the model characterises the binocular disparities that are likely to be encountered in the natural environment, and furthers our understanding of binocular visual systems.