Model adaptive image restoration

A new model adaptive method is proposed for restoration of blurred and noise corrupted images. This approach exploits information available from observed data to choose the appropriate optimization criterion and produce an approximate maximum likelihood solution. The generalized p-Gaussian family of probability distributions is used to model a wide range of observed noise classes. Distribution shape parameters are estimated from the image, and the resulting maximum likelihood optimization problem is solved. A fast iterative algorithm for this method is presented and analyzed. Experimental results indicate that this method outperforms the least squares method by taking advantage of the non-Gaussian characteristics of the noise data.<
>