Spatial Correlation Analysis of Polarimetric SAR Data by Means of the Structure Tensor

Abstract:
This paper proposes a new approach for the analysis and characterisation of nonstationary spatial texture in polarimetric SAR data. A previously introduced model named Anisotropic Gaussian Kernel (AGK) was shown to be a pertinent indicator of local texture orientation and permitted to represent in a simple way the complex spatial structure of SAR intensity. We propose here two methods for AGK parameter estimation. The first one is an enhancement of a previously developped algorithm and the second one is a new approach based on the Gradient Structure Tensor (GST). These two methods are applied to PolSAR data.