SIMULATED ANNEALING - IMAGE-RESTORATION AND SURFACE RECOGNITION

This article follows the one of C. Bonnemoy and S. B. Hamma on the practice of simulated annealing in global optimization in R(n) (cf. [BON91a]). One presents now two applications: the first, classical, about image restoration, where the Metropolis dynamics have been choosed instead of the Gibbs sampler. The second, new, for the shape from shading problem. In the first case the state space is finite and is of the form L(s) where L is the space of grey levels and S is the pixels network. In the second case, the state space is continuous and is of the form R(s) where S is a finite network in R2. This second case is, for the convergence, related to a theorem which can be used for the optimization problem in R(n), recalled at the beginning of the present article.