Toward global solution to MAP image restoration and segmentation: using common structure of local minima

摘要

In this paper, an iterative optimization algorithm, called the Comb algorithm, is presented for approximating the global solution to MAP image restoration and segmentation. The Comb derives new initial configurations based on the best local minimum found so far and leads a local search towards the global minimum. Experimental comparisons show that the Comb produces solutions of quality comparable to simulated annealing.