Non stationary Bayesian image restoration

We propose a new iterative Bayesian non stationary image restoration algorithm. The main novelty of this approach is the introduction of a hierarchical non stationary image prior. Based on this prior and the generative graphical model for the observations, Bayesian inference is performed integrating out the hidden variables. An interesting byproduct of this approach is the justification, using a Bayesian framework, of previous non stationary image restoration formulations that were based on heuristic arguments. Numerical experiments are provided that demonstrate the advantages of the proposed non stationary approach as compared with the stationary approaches.