Adaptive Speckle Filtering of SAR Images Using Wavelet-based Method

Abstract

Since speckle appearing in SAR images as a granular noise must be eliminated for the best image classification and interpretation. This paper presents a technique for despeckling by using the stationary wavelet decomposition to detect edge and non-edge regions. The edge regions can be detected from three directional wavelet coefficient energies. The large wavelet coefficients are generated by edge and then removed small wavelet coefficients or non-edge regions by using various filtering in all high-frequency bands. The despeckled image is obtained by reconstruction from the filtered coefficients. Experimental results on JERS-1/SAR images showed that the proposed method gives a significant noise removal and preserves the sharp features of the images, as well as visual appearance quality.