A Statistical Polarimetric Decomposition Solution Based on the Maximum-Likelihood Estimator

This letter addresses a statistical model-based decomposition solution for polarimetric synthetic aperture radar imagery. The Wishart distribution is introduced to the two-component Freeman-Durden (2FD) model to enhance the traditional direct solution (2FD-DS) accuracy. This letter proposes a maximum-likelihood estimator (MLE) (2FD-MLE) expression which is simple enough to numerically solve 2FD unknowns. Furthermore, the statistical randomness impact is observed for the first time. The authors go on to verify that the decomposition results can be greatly improved by MLE, even in a simple physical model. The experiments show that the MLE enhances the estimation accuracy of land-cover types. At a moderate-look scale, the 2FD-MLE has less negative span flaws than the 2FD-DS method, and the estimation results are more close to the physical interpretation.