Anisotropic diffusion techniques on synthetic aperture radar data

Abstract

Speckle in SAR imagery is a by-product of constructive and destructive interference between scatterers within a resolution cell. This speckle phenomenon gives SAR imagery a noise-like appearance and is often exploited in near angle and/or coherent stereo pairs. However, in many cases, this speckle is unwanted and can be considered noise or interference. We use partial differential equation (PDE) methods for speckle mitigation in detected imagery and the collected complex image data. In particular, we study the effects of non-linear anisotropic diffusion filters on collected SAR image data. In the past, anisotropic diffusion (AD) techniques have been successfully used in the analysis of EO data. However, the use of these techniques on SAR image data is recent and much is yet to be done. We expect the application of AD techniques on SAR image data in combination with a fluid dynamic perspective to yield rich dividends in terms of image interpretability. Through our approach we demonstrate that it is possible to spatially maintain areas of high dynamic range (bright scatterers) and smooth areas of low dynamic range in the scene. We also exhibit the role of these non-linear filters in correlation, registration, compression, decompression, and image interpretability for SAR analysts.