Robust Covariance Matrix Estimation with Application to Volcano Monitoring using SAR Image Stacks

Abstract:
This paper introduces a new method - the Rank M-Estimator (RME) - for robust covariance matrix estimation of distributed scatterers (DSs) in SAR image stacks, or more general for complex multivariate with multiplicative and non-stationay phase signal. The RME can work without the assumption of samples’ stationarity, which is seldom addressed in the SAR community. In other words, no flattening/estimation of the interferometric phase is required. The robustness of RME is achieved by using an M-estimator, i.e. amplitude-based weighing function in covariance estimation.