A CHMT model based DE-speckling method for SAR image

A coarse-classification based tying method for the Contourlet-domain Hidden Markov Tree model (CHMT) solution algorithm is proposed to speed up the parameters estimation; and a general SAR image filtering framework, to which any kind of shift-variant transform can be applied, is generated by applying together with the LOG Transform, mean rectification and cycle-spinning, etc. The proposed coarse classification based tying method for CHMT is applied to de-speckle the SAR image in the general framework, and the result is compared with those of some commonly-used filters. The visual effects and the statistical parameters indicate that the coarse-classification based tying method for CHMT is much faster than the other tying methods, and the CHMT based de-speckle method can achieve better result than some commonly-used filters.