基于子矩阵对数累积量的极化合成孔径雷达图像等效视图数估计新方法
摘要
The equivalent number of looks is an important parameter in the processing of polarimetric synthetic aperture radar (Pol-SAR) imagery. In this paper, novel estimation methods named maximum likehood (ML) estimator and union ML estimators are presented based on log-cumulants of the sub-covariance-matrices. The novel methods can not only be applied to the statistical model of Gaussian, but also to product model. Then the performance of different ML estimators with sub-covariance-matrices of different dimensions is analyzed, and the comparisons among the novel and existed methods including the Cramer-Rao Bound (CRB) are given. The statistics of the determinant of the covariance matrix is given to improve the estimation accuracy. At last the correctness and effectiveness of the novel methods are verified via simulated data and real data, especially for the product model, which really improve the rapidity and accuracy of parameter estimation in Pol-SAR imagery.