SAR Images Despeckling Based on Hidden Markov Mixture Model in the Wavelet Domain
Yan Xiang Wu, Qiang Zhang, Xia Wang, Guisheng Liao · 2006
In this paper, an efficient despeckling algorithm is proposed based on the hidden-state Markov random field (MRF) and the hidden Markov tree (HMT) in the wavelet domain for synthetic aperture radar (SAR) image. The minimum mean square error (MMSE) despeckling technique without the log-transform is fused in the algorithm. This algorithm also employs a new hidden Markov half tree model, which improves its computational speed. The clustering and the persistence of wavelet coefficients are taken into account in this model, which are characterized by the MRF model and the HMT model respectively. Experimental results show that our method achieves good performance in terms of noise suppression and edges preservation, and that its running time is less than that of the HMT by twenty times approximately