Image restoration based on wavelet-domain contextual hidden Markov tree model

From the viewpoint of Bayesian method, image restoration algorithms based on wavelet-domain hidden Markov tree (HMT) model have been proposed recently. These algorithms utilize the HMT model which captures the persistence property of wavelet coefficients, but lack the clustering property of wavelet coefficients within a scale. In this paper, we propose a new image restoration algorithm. The algorithm specifies the prior distribution of real-world images through wavelet-domain contextual hidden Markov tree (CHMT) model which enhances the clustering property of the HMT model by adding extended coefficients associated with wavelet coefficients and converts the restoration problem to a constrained optimization task. Experimental results show that, the proposed algorithm produces almost better results than the HMT model produces for image restoration, both in objective and subjective qualities.