Genetic Blind Image Restoration With Dynamical Local Evaluation

The blind image restoration problem consists in estimating the original image from blurry and noisy data, without knowing the involved blur operator. The problem is well known to be ill-posed even in the not-blind formulation, nevertheless the use of regularization techniques allows to define the solution of the problem as the minimum of an energy function. In this paper we solve the blind restoration problem with a evolutionary approach. A population of blur operators is evolved with a fitness given by the opposite of the energy function to be minimized. Since the fitness evaluation, calculated on the whole image, represents a significant computational overhead which can make the method unfeasible for large images, an original technique of dynamical local fitness evaluation has been designed and integrated in the evolutionary scheme. The subimage evaluation area is dynamically changed during evolution of the population. The underlying hypothesis is that the explored subareas are significatively representative of the features of blurs and noises in the global image. The experimental results confirm the adequacy of such a method: in some cases the proposed genetic blind reconstruction finds qualitatively better solutions outperforming the not-blind standard deterministic algorithm.