Optimum parameter estimate for K-distributed clutter using multiple moments

The authors analyze the sub-optimal performance of simple texture measures for estimating the reciprocal order parameter of K-distributed radar clutter. A non-committal neural net has been applied to the parameter estimation task which has shown that improved error estimates are obtained when multiple moments are used to characterize the texture. Prompted by this result a new estimator is proposed which combines the mean normalized log intensity and the amplitude contrast moments of the imaged data. Its error performance is determined by the relative weighting in which the two moments are combined. With an appropriate choice of the weighting the modified estimator outperforms the normalized log estimator and gives close to maximum likelihood performance on the estimates over a wide range of the parameters values which are of interest.