Model-based processing of multifrequency polarimetric SAR images of urban areas

In this paper, we describe a two-step classification scheme for fully polarimetric SAR images. The classification scheme is composed of the cascade of an optimum segmentation stage, and an ML supervised classifier. Different segmentation schemes are described, specifically designed for mono- or multifrequency images. The classification scheme is applied to a set of fully polarimetric, multifrequency SIR-C images of the town of Pavia, in Northern Italy, considering all the possible pairs of polarimetric channels and the two bands individually and jointly, aiming at identifying the best combination for practical applications. Results show that for urban areas, the best performance is achieved by jointly processing the three polarimetric channels, and the minimum performance degradation is achieved considering the HH and the HV channels.