SAR Image Classification Using Non-Parametric Estimators of Shannon Entropy

Abstract

Remote sensing with Synthetic Aperture Radar (SAR) data has become a vital tool for environmental studies. But they suffer the presence of speckle noise which hinders the analysis and interpretation of this kind of images. It has been shown that the ${\mathcal{G}^0}$ family of distributions is a suitable model for SAR intensity because it possesses the ability to characterize areas with different degrees of texture. Information theory has gained a place in signal and image processing for parameter estimation and feature extraction. In this paper we evaluate the performance of different parametric and non-parametric Shannon entropy estimators in the single look cases, as attribute in the algorithm used for classification SAR images. These estimators were analyzed through Kappa and accuracy indexes. Finally, we apply these estimators to actual data.