Intelligent Wavelet Coefficients Thresholding: Speckle Reduction Approach in SAR Imagery

Synthetic aperture radar imagery contain valuable information that are used in most cases, in this regard reducing the pseudo-noise effect of speckle seems to be necessary. This research aimed to assemble a hybrid structure to deal with the aforementioned challenge based on the indirect intelligent thresholding of wavelet decomposition coefficients named intelligent wavelet coefficient thresholding. In order to automatically determine the threshold limit, a hybrid model was used, in which the wavelet transform coefficients were searched using the combination of meta-heuristic optimization methods of genetic and particle swarm optimization. Unlike previous researches, which mainly used the cost function based on the least square error and the signal-to-noise level, in this research, the cost function of the total variation was used. The proposed algorithm was applied to both single-look images and multi-look images. Compared to adaptive spatial filters and statistical thresholding of wavelet coefficients, the proposed algorithm of analytical indices of speckle reduction performance increased the equivalent number of look and edge preservation index by 10 and 2.5 times for multi-look images and 10 and 5 percent for single-look images, respectively. Based on the results, proposed method which suggests indirect wavelet coefficients thresholding, performs well in speckle reduction.