The inadequacy of Gaussian probability models in describing certain regions of a SAR image is explained by the violation of some fundamental Gaussian assumptions. This can be attributed to the increase in spatial resolution of recent sensors as well as target heterogeneity. This project will look into non-Gaussian probability models, like the G distribution, for single-channel and polarimetric (i.e. multivariate) SAR data and will propose a new framework for its parameter estimation.
Synthetic aperture radar (SAR) data are inherently probabilistic due to the presence of speckle, which is a characteristic phenomenon of coherent imaging. Speckle appears un-ordered, random, and noise-like. This hampers SAR image analysis, and mandates statistical modelling to form a fundamental step in information retrieval algorithms like classification, segmentation, target identification, and change detection.