Parameter estimation for compound‐Gaussian clutter with inverse‐Gaussian texture

Abstract

This study deals with the estimation problem of the inverse-Gaussian (IG)–compound-Gaussian (CG) distributed clutter parameters. Under the assumption of the absence of thermal noise, non-integer moments, log-moments and maximum-likelihood-based approaches are employed for model parameter estimation in the cases of single pulse and non-coherent integration of N pulses. The forms of the non-integer-order moments estimator and the [z log(z)] estimator are derived in terms of the Bessel and the exponential–integral functions, respectively. These formulas maintain monotonic nature over all values of the ratio between the shape parameter and the mean clutter power. For a single look data, the ML estimate combined with the mean clutter power is constructed in one dimension search for the shape parameter. By accommodating the mean square error metric, the estimation assessments are investigated via simulated and real data. Authors’ results illustrate that the derived non-integer-order moments estimator is asymptotically efficient for the IG–CG distributed clutter parameters particularly in heavy tailed clutter situations.