SAR De-speckling using a Combination of Thresholding and MMSE/MAP Bayesian Estimation in Contourlet Transform Domain

In the receiver antenna of SAR system electromagnetic waves backscattered from the target surface, add together coherently and random interference of these waves causes the speckle noise in reconstructed SAR images. Speckle may be modeled as a multiplicative noise which degrades the quality of SAR images. In this paper, we first present a review of low-complexity and high-speed despeakling algorithms which are developed based on thresholding and Bayesian estimation in non-subsampled contourlet transform (NSCT) domain. In usual thresholding methods when input amplitudes fall below a given threshold they are mapped to zero, however such amplitudes may contain useful information about image details. In order to simultaneously gain the low-complexity property of thresholding and high precision of Bayesian estimators, we then suggest a new thresholding method in which small-amplitude NSCT coefficients are estimated using LMMSE/MAP filters. The performance of despeckling filters is quantitatively evaluated on both simulated data and real SAR image using statistical indexes. The results illustrate the superior performance of the proposed algorithm in comparison with usual thresholding and Bayesian MMSE/MAP filters.