A hierarchical method of MAP-based stochastic diffusion and disparity estimation

This paper talks about a hierarchical approach on stochastic diffusion in the MAP-based estimation. Stochastic diffusion has been proposed for an optimization method to minimize the potential function in the MAP-based estimation, and showed good performances in the simultaneous estimations of correspondence, line, and segmentation fields. This paper applies stochastic diffusion to the MAP-based estimation of disparity and line fields in the hierarchical scheme. The proposed hierarchical method combines two successive approximations of the disparity field and potential space at the same time. This hierarchical method propagates not only the geometric relation but also interactions of neighborhood fields. The experimental results show that the proposed hierarchical stochastic diffusion decreases the memory and computational burden and improves the estimation performances in the occluded or textureless regions.