Analysis of SAR speckle statistics in support of image filtering and interpretation

Abstract

This study presents an automated approach towards the evaluation of speckle noise statistics in SAR images. Based on the statistical information we derive a textural feature – the so-called speckle divergence - which describes the deviation of local speckle behavior from the scene-specific heterogeneity of fully developed speckle. We show that the speckle divergence is a valuable input parameter for adaptive speckle suppression by introducing a spatially adaptive and radiometrically selective speckle filter algorithm. Moreover the benefit of speckle divergence for both the identification of built-up areas and the generation of a color composite product on the basis of one single-polarized SAR image is presented.