Convex relaxations for intractable variational approaches have played a major role in recent years in the fields of mathematical imaging and computer vision. Concerning shape in connection with image segmentation, however, approaches are still lacking that meet all desired requirements: invariant matching and unique assignments, independence of initializations, statistical shape prior knowledge, low runtime complexity. The talk will review recent progress on related subproblems with a focus on the overall problem.
Copyright. All Rights Reserved.