SAR image despeckling based on wavelet kernel transform and Gaussian scale mixture model
Fan Liu, Licheng Jiao, Shuyuan Yang · 2009
A new method about SAR image despeckling is proposed in this paper, this method is achieved by combining wavelet kernel transform (WKT) and Gaussian Scale Mixture model (GSM). WKT is a multiscale transform which is based on machine learning model. By analysis the distribution of the coefficients after WKT, these coefficients are similar to Gaussian distribution, and these noised coefficients are distributed as Gaussian too, but are independence with non-noised coefficients. In this paper, we construct the neighbor model based on the coefficients after WKT, and use the Bayes least mean square to despeckle the spots in SAR images, and the model describes the edge distribution of coefficients. We use the proposed method to process the SAR images, and the results demonstrate that this method can obtain better denoising images than Lee filter, wavelet transform etc.