Innovative Solutions Based on the EM-Algorithm for Covariance Structure Detection and Classification in Polarimetric SAR Images

This article addresses the challenge of identifying polarimetric covariance matrix (PCM) structures associated with a synthetic aperture radar (SAR) image. Interestingly, such information can be used, for instance, to improve scene interpretation or enhance performance (possibly PCM-based) segmentation algorithms as well other kinds methods. To this end, general framework solve multiple hypothesis test is introduced aim detect and classify contextual spatial variations in SAR images. Specifically, under null hypothesis, only one unknown structure assumed data belonging two-dimensional sliding window, whereas each alternative are partitioned into subsets sharing different PCM structures. The problem partition estimation solved by resorting hidden random variables representative classes expectation–maximization algorithm. effectiveness proposed detection strategies demonstrated on both simulated real also comparison existing classification algorithms.