Coherent Change Detection for Multi-Polarization SAR

This paper presents a solution to the coherent change detection (CCD) problem using multi-polarization synthetic aperture radar (SAR) imagery. The multi-polarization SAR imagery (i.e., the day-1 reference and day-2 test images) are modeled as jointly correlated complex Gaussian vectors with unknown correlation, rho = gammae^jPhi . Maximum likelihood estimates of the unknown phase and coherence parameters (Phi,gamma) are derived. Accuracy of the MLE estimates is evaluated; and the benefit of using multi-polarization data versus single-polarization data is quantified