ESTIMATION OF TEXTURE PARAMETERS IN K-DISTRIBUTED CLUTTER

The authors analyse some of the problems associated with estimating the order parameter of K-distributed high-resolution radar clutter. They identify the cause of these problems to be the nonlinear inversion of various texture measures to the order parameter of the texture. Accordingly, they propose a reciprocal parameter to describe the order. They derive the resultant systematic and random errors in the new texture descriptor, and demonstrate that this removes many of the problems. Simulations now show agreement with the predicted bias and variance, while the bias is significantly smaller than before. It is also confirmed that the normalised log texture measure remains more accurate than others which have been proposed, and approximates the maximum likelihood solution to the estimation of the order parameter over most of the range of interest.