Turbo iterative estimation of singularity structure in SAR image based on wavelet-domain hidden Markov models

Wavelet-domain hidden Markov models (HMMs) have been widely applied to image processing, e.g., image restoration. The models provide great promise of detecting image singularity structures with some hidden states. However, these hidden states are rather difficult to estimate, especially under the influence of the multiplicative speckle noise in SAR images; no efficient estimation method has been developed yet. By using the principle of turbo iterative decoding, we propose a new turbo iterative method to estimate the hidden states of the wavelet-domain HMMs for SAR images. In our method, hidden states are estimated alternatively in two orthogonal sub-spaces with a soft estimation scheme, and the posterior probability is exchanged between the two subspaces. The experimental results of the proposed method illustrate rather an impressive performance.