Statistical methods for the evaluation of the coefficient of variation in SAR images

Abstract

The speckle noise present in SAR images has a multiplicative nature, with unity mean and standard deviation (Std) /spl sigma//sub n/. The authors introduce two new techniques that estimate its statistical properties. The proposed techniques estimate the coefficient of variation (R), using local statistical properties which characterise homogeneous areas in SAR images. Accurate estimates of R are often obtained by methods based on the manual extraction of homogeneous blocks from an image. In the proposed techniques the selection is made automatically using statistical properties of the speckle.