Improved recursive inverse filtering method for blind image restoration

For inverse filtering methods of blind image restoration, great emphasis has been put into considering the global features of images. Local space statistics of the observed image are estimated recursively and integrated to approach those of the original image. These estimated statistics are refined iteratively and gradually converge to the true values. Combined with the technique of nonnegativity and support constraints recursive inverse filtering (NAS-RIF), an improved recursive inverse filtering algorithm is proposed. Simulations show that the new regularized method behaves better in respect of computational stability, noise resistance, and edge preservation ability.