Polarimetric SAR Data Decomposition Based On Polarimetric Signatures And Reference Scattering Models

The polarimetric decomposition is one of the most important steps in SAR data processing and analyzing. Conventional decomposition methods use polarimetric information only in a restricted number of polarization bases. This paper presents a new decomposition method based on polarimetric signatures. The proposed decomposition includes two main steps: 1) selection of the reference polarimetric signatures, and 2) Classification of the pixel's polarimetric signature. The presented method was tested on the Radarsat-2 image in C band collected over San Francisco and the Pi-SAR image in L band collected over Niigata University in Japan. The proposed decomposition was compared with Y4O, Y4R, Arii-NNED and Freeman decomposition methods. According to the results of the suggested method, in urban areas especially in areas with a large orientation angle, it is clearly seen that the overestimation of the volume contribution has been reduced and the double-bounce contribution has been increased. In addition, the obtained power values of the proposed decomposition are always positive; therefore, the problem of negative power value cannot be achieved.