A statistical method for discrimination of natural terrain types from SAR data

We propose a method for discrimination of underlying textural structures from spotlight-mode synthetic aperture radar (SAR) returns by using a tomographic data acquisition model as the basis for statistical reasoning. We model the hypothesized textures by statistically self-similar processes and formulate the problem in a hypothesis testing framework in the SAR range profile domain without any image formation. We achieve a near-optimal, computationally efficient evaluation of the likelihood test by transforming the data into the multiscale domain