Sparse Bayesian blind image deconvolution with parameter estimation

In this paper we propose a novel blind image deconvolution method developed within the Bayesian framework. A variant of the non-convex l_p-norm prior with 0 <; p <; 1 is used as the image prior and a total variation (TV) based prior is utilized as the blur prior. The proposed method is derived by utilizing bounds for both the image and blur priors using the majorization-minimization principle. Maximum a posteriori Bayesian inference is performed and as a result, the unknown image, blur and model parameters are simultaneously estimated. We also show that as a special case, the developed method provides very competitive non-blind image restoration results when the blurring function is assumed to be known. Experimental results are presented to demonstrate the advantage of the proposed method compared to existing ones.