Smoothing speckled SAR images by using maximum homogeneous region filters: an improved approach

Abstract

Because the speckle phenomenon corrupts the visibility of SAR images, many techniques have been proposed to improve the data. In this paper, we present an improved first step in the filtering process by using adaptively tailored windows in order to select maximum homogeneous regions. This selection is based on a classical growing region method tuned by the variations of the local estimate of the equivalent number of looks L: dealing with an analytical expression for the variance of L-estimator, the proposed segmentation seems to be more realistic than the initial method.