Segmentation via annealed motion estimates

A segmentation and motion estimation are two problems that require accurate estimation for many applications in computer vision and image analysis. In this paper, an algorithm that is based on motion information and restored spatial information to segment the scene is presented. This algorithm has minimum dependency on noisy data and the motion vectors are estimated using a deterministic mean field annealing (MFA) framework. The algorithms results are very good using synthetic and real world sequences.