Region tracking using level set registration

We present a technique for region tracking using a novel method of registration. This method uses the representation of the image with a tree form containing the topological and geometrical information in the image. The computation of the tree is achieved by using the so called Fast Level Set Transform (FLST) and owns the fine property of contrast invariance. We explain the main reasons why the shapes present in the image correspond to some nodes in the tree. We briefly describe the basic principles of the FLST. We then expose a registration algorithm, based on a vote procedure, that consists of finding correspondences between the nodes of the trees of two different images. We also porpose an implementation of this algorithm for a real time application and end with some numerical experiments.© (2001) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.