Speckle Reduction of SAR Images for Small Feature Extraction

Abstract

A new, quantitative and consistent measure of contrast is proposed as a way to evaluate the preservation of small features in SAR images. A despeckling filter should reduce speckle yet preserve very small features of interest to facilitate the subsequent automatic extraction of very small or thin features from a SAR image. With this measure and the speckle index defined by Lee et al., we evaluated 9 despeckling filters using both 19 bright and 19 dark features with varying intensities and contrasts in observed images. We found that the SFP filters developed by the authors is best for extracting bright features. For dark features, the Sigma filter is best, with special situations in which the Enhanced SFP filter performs better.