Blind deconvolution using maximum a posteriori (MAP) estimation with directional edge based priori

摘要

Image inverse problem or ill-posed problem has been tackled by various researchers in a number of ways. The number of unknown parameters as well as the properties of the parameters makes the problem very challenging. Majority of the research has been focused on using image properties and any prior knowledge of the bwr or degradation phenomenon. The whole process can be classified in to three categories: (i) blind methods; (ii) semi blind methods and (iii) non-blind methods. The proposed work in this paper is based on semi blind method where partial prior in the form of directional edges were used. The Bayesian class of methods using maximum a posteriori (MAP) estimates were proven very successful and used here. The main contribution of the proposed method involves (i) alternating minimization step for PSF (blur) and shape image; (ii) regularization based on adaptive directional gradients and (iii) non-local minima using edge features. The experimental result shows that the proposed algorithm is better than other enriching methods. (C) 2017 Elsevier GmbH. All rights reserved.