Efficient Computation of Closed Contours using Modified Baum-Welch Updating

We address the problem of computing closed contours delineating the boundaries of objects of interest from image tangent maps. The Bayesian formulation of this problem incorporates the notion of foreground and background models in inferring the optimal image organization. Due to vital non-local constraints, the optimal MAP solution is intractable. We present a novel efficient search algorithm for the construction of highly probable closed contours, based on the Baum-Welch algorithm for updating hidden Markov model parameters.