THE MODIFIED BETA DENSITY-FUNCTION AS A MODEL FOR SYNTHETIC APERTURE RADAR CLUTTER STATISTICS

In generating automatic classification algorithms from SAR image data it is important to have an adequate model for the underlying distribution function of the random variable used to model the image data. A number of different candidate distribution functions have been presented previously, but no single model has been shown to represent a range of clutter targets. We show here that the modified beta distribution function is such a model that can adequately represent a range of SAR returns from different ice types by a simple change in its parameters. This ability is explained by describing the distribution functions in width, modified skewness space where the modified beta function covers a region, while the other, more common, distribution functions cover only a curve. A procedure for comparing sample distribution functions with analytical functions specifically for digitized SAR data is presented, and the modified beta model is tested on 166 subsets drawn from three SAR collections over different ice types and over open water. We show that the modified beta function can model essentially all of the SAR subsets, where the other more common densities cannot. Finally, we present some classification of ice types using the parameters from the modified beta function.