NON-LOCAL SAR IMAGE DESPECKLING BASED ON SPARSE REPRESENTATION

Speckle noise is an inherent problem of synthetic aperture radar (SAR) images, which not only affects the acquisition of SAR image information, but also increases the difficulty of image interpretation and analysis, and greatly reduces the efficiency of image segmentation and feature classification. Based on the nonlocal idea of SAR-BM3D and the concept of sparse representation, a new synthetic aperture radar (SAR) image speckle noise suppression algorithm is proposed. The algorithm follows the framework structure of SAR-BM3D, performs logarithmic conversion after the coarse filtering in the first step, replaces the three-dimensional similar block group after the similar block matching in the second filtering step with a two-dimensional matrix stacked by columns, and replaces the local linear minimum mean square error (LLMMSE) denoising with K-SVD sparse representation image denoising for speckle noise suppression of SAR images. The experimental results show that the non-local denoising method based on sparse representation achieves better results in terms of evaluation metrics and visual perception quality compared with several existing advanced reference techniques.