Performance Analysis of Multivariate Super-resolution Processing of Polarimetric Synthetic Aperture Radar Tomography

Abstract

In a previous paper, the authors developed a technique known as Vector Linear Prediction (VLP) which achieved super-resolution processing for polarimetric SAR tomography. In this paper, the performance of this multivariate super-resolution processing technique is investigated. Both simulations and field data are used to assess the limitations of this technique due to the loss of spatial and polarimetric coherence. The model assumes two scattering centers in each image pixel. The study shows that if one component remains linearly polarized, while the other is fully depolarized, then meaningful interferometric information can still be retrieved for the linearly polarized scattering center. However, if the depolarized component is also spatially decorrelated (loss of spatial coherence), then the interferometric phase estimates of both components are prone to significant errors. The field data is used to verify and validate the observations obtained with the simulations.