Wavelet Domain Image Restoration and Parameters Estimation based on Variational Bayesian Method and Student-t Priors

In this paper we proposed a novel algorithm for image restoration and parameters estimation with a wavelet domain Student-t priors using variational Bayesian method. Because the conventional spatial domain hierarchical prior models have the shortcoming that normalization constant cannot found in closed form, we resort to wavelet domain prior model that avoid the above shortcoming. The restored image and the unknown hyper-parameters and parameters are simultaneously estimated with variational expectation maximize (EM) algorithm. Numerical experiments are presented that demonstrate the advantage of this algorithm as compared to other image restoration algorithms.