Disparity Using Feature Points in Multi Scale

Ilkay Ulusoy, Edwin R. Hancock, Ugur Halici

Abstract: In this paper we describe a statistical framework for binocular disparity estimation. We use a bank of Gabor filters to compute multiscale phase signatures at detected feature points. Using a von Mises distribution, we calculate correspondence probabilities for the feature points in different images using the phase differences at different scales. The disparity map is computed using the set of maximum likelihood correspondences.