Estimation of the degree of polarization of coherent light in the presence of uniform and nonuniform illumination
Résumé
We address estimation of the degree of polarization (DOP) in active images acquired under coherent light when the coherency matrix of the backscattered light is diagonal. We consider cases where the illumination intensity is uniform on the scene and when it is not, and we take into account different parametrizations of the DOP. In all cases, we determine the Maximum Likelihood (ML) estimators and compare their variances with the Cramer-Rao Lower Bound for unbiased estimation. When illumination is not uniform, the ML estimators are solutions of equations that have to be solved numerically. We show that simpler estimators based on the trimmed mean make it possible to approach performance of ML estimators at lower computational cost.