Iterative adaptive lp restoration of blurred images

A new model adaptive method is presented for restoration of blurred and noise corrupted images by exploiting information available from observed data to choose the appropriate optimization criterion. The derived maximum likelihood solution is based on the l_p minimization criterion that naturally arises from the adoption of the generalized p-Gaussian family of probability distributions as an additive noise model. A fast and efficient iterative algorithm for this adaptive method is developed and analyzed. Experimental results indicate that this method adapts to the non-Gaussian nature of the noise process and outperforms the least squares method, which lacks the flexibility of the former method