Skip to Main Content
Wavelet-domain hidden Markov models (HMMs) have been recently proposed and applied to image processing, e.g., image denoising. In this paper, we develop a new HMM, called hierarchical hidden Markov tree model (HHMT), by adopting a feasible and fast two stage algorithm which avoids the time-consuming training process to estimate the HMT model parameters. The HHMT can exploit both the local statistics and the interscale dependencies of wavelet coefficients at a low computational complexity. We show that the HHMT model can achieve state-of-the-art medical image denoising performance.