Speckle filtering of SAR images using Hölder regularity analysis of the sparse code

This paper introduces a novel approach for filtering speckle in SAR images using the Hölder regularity analysis of the sparse code. Recently, sparse code shrinkage was proposed for filtering images corrupted with additive white Gaussian noise. Sparse coding is a method for finding a representation of the data in which only a small portion of the basis is simultaneously active. In the proposed approach, first the sparse code representation for the SAR image is obtained. Then, the pointwise Hölder exponent is computed for each of these basis functions and used as a cost function for basis enhancement. This enhancement is the key for reducing speckle in SAR images. As the results show, the reconstructed images have a better visual perception as compared to those obtained by adaptive filtering techniques and coefficient shrinkage based algorithms.