Model-based decomposition with adaptive selection of unitary transformations

In this paper a three component model-based decomposition with adaptive selection of unitary transformations for polarimetric synthetic aperture radar (POLSAR) data processing is proposed. Singh et al implemented two unitary transformations on the coherency matrix to minimize the power of cross-polarization, and as a result the T₂₃ element of the coherency matrix becomes zero. Another two unitary transformations are proposed by us to carry out on the coherency matrix also to minimize the power of cross-polarization, and the T₁₃ element of the coherency matrix becomes zero. Here, we first implement Singh´s two unitary transformations and the proposed two unitary transformations on the coherency matrix separately. Then we select the one which leads to the smaller T₃₃. At last, we carry out the three component model-based decomposition proposed by Freeman and Durden based on the obtained coherency matrix. The smaller is obtained, the better the over-estimation of volume scattering in model-based decomposition can be suppressed. The RADARSAT-2 POLSAR data of San Francisco area is used to validate the improvement of the proposed method over the three component decomposition only with Singh´s two unitary transformations.