Super-resolution image reconstruction based on wavelet-domain classified hidden Markov tree model

From the viewpoint of Bayesian method for image reconstruction, a super-resolution algorithm based on the wavelet-domain classified hidden Markov tree (CHMT) model is proposed. The CHMT model is used as a priori information of the image in the wavelet-domain. The distribution densities of the wavelet coefficients probabilities can be approximated by the Gaussian mixture model. And the reconstruction problem is converted to a constrained optimization task, which can be solved by the conjugate gradient method. The method to adaptively determine the regularization parameter is also proposed. Experimental results show that the proposed algorithm has a reasonable computational complexity, and both the PSNR and the subjective visual effect of the reconstructed image are improved.