New phase-difference for polarimetric SAR images

Classification of urban areas in SAR images has always been a difficult task [1]. For urban area, however, backscattered signals in general are from single-bounce, double-bounce and multiple-bounce returns of various parts of the structure, although they are not from the defused scattering of a random media such as vegetation [1]. The orientation angle of scattering media affects the polarimetric radar signatures [2]. So Lee et al. [2] and Yamaguchi et al. [3] consider transformation of coherency matrix T3 to model rotations about the line-of-sight. In this paper, we will present the phase difference characteristics of urban areas in polarimetric SAR (PolSAR) images and an improved classification algorithm for urban areas using the co-polarized phase difference (CPD) and cross-polarized phase difference (XPD).