a test statistic based on wishart distribution for unsupervised change detection in multilook polarimetric sar data

in this paper, an unsupervised method for change detection in the multitemporal multipolarization synthetic aperture radar (sar) imagery is proposed. a matrix distance measure, named symmetric revised wishart is used as a test statistic in order to assess the similarity of two multitemporal multilook polarimetric sar data and a variance-based thresholding algorithm is applied to the test statistic image to obtain the final optimum change map. experimental results, which confirm the accuracy of the proposed method for simulated polarimetric sar data and radarsat-2 spaceborne imaging radar c-band images, are presented. the findings indicate that 20.3% higher detection accuracy is obtained using the multilook full polarimetric covariance data than the diagonal elements case of covariance matrices.