A comparative study of sea clutter covariance matrix estimators

Abstract

Estimation of the polarimetric covariance matrix is an important task in statistical modeling of sea clutter for maritime applications of polarimetric synthetic aperture radar (PolSAR) data. This work provides a comprehensive study of four covariance matrix estimators: the maximum likelihood estimators under the Gaussian distribution (G-ML) and the K distribution (K-ML), an approximation of the latter (AK-ML), and a robust M-estimator. It adds to previous theoretical studies of these algorithms by evaluating their performance with respect to both estimation accuracy and computational efficiency. Experiments are performed on simulated datasets. Various texture conditions of the sea clutter are considered in the study.