Improved anisotropic diffusion filtering for SAR image despeckling

Abstract

In synthetic aperture radar (SAR) images, the speckle noise often corrupts salient information that is of interest (e.g. textures, small hard targets, object boundaries, etc.). To tackle such an issue, proposed is a novel anisotropic diffusion filter that manages to simultaneously fulfil competing requirements: noise reduction on homogeneous regions, weak edges preservation, and keeping hard targets intact. The capabilities of the proposed filter were proved by comparing it with another two non-linear diffusion filters applied on a Lena image corrupted by a multiplicative speckle noise and on a real SAR image acquired by the Cosmo-SkyMed satellite constellation.