Wavelet-based image denoising using hidden Markov models

Wavelet-domain hidden Markov models (HMMs) have been recently proposed and applied to image processing, e.g., image de noising. In this paper, we develop a new HMM, called local contextual HMM (LCHMM), by introducing the Gaussian mixture held where wavelet coefficients are assumed to locally follow the Gaussian mixture distributions determined by their neighborhoods. The LCHMM can exploit both the local statistics and the intrascale dependencies of wavelet coefficients at low computational complexity. We show that the proposed LCHMM combined with the "Cycle-spinning" technique may achieve the best performance in image denoising.