Bayesian approach to SAR image reconstruction

An approach for reconstruction of speckled SAR images is presented. This approach is based upon Bayes´ rule obtaining the maximum a posteriori estimate of the underlying radar cross section. The prior used for the reconstruction is modelled by Gibbs random fields reflecting the existing texture characteristics, while the system transfer function of the SAR signal processing together with the speckle noise is accounted for by the likelihood distribution. The solution of this optimization problem is obtained by relaxation methods such as simulated annealing using the prior information as a constraint to limit the optimization space