Statistical and phenomenological recognition in polarimetric SAR imaging

The maximum likelihood classifiers are of large interest because they allow one to simulate, quantify and compare easily the rate of correct recognition for various cases of partial polarimetries and symmetries hypothesis. In this paper, they are used within the framework of natural surfaces recognition from SAR polarimetric images. The purpose is to advise the choice of the more suitable partial polarimetry for a remote sensing satellite. The results of this work infer some questions about the validity of some properties often supposed in the analysis of SAR images and in the radar calibration method.