The detection of weak signal patterns in radar ocean intensity images

Abstract

The detection of weak patterns in radar ocean RCS images is complicated by the fact that signals and noise are interactive rather.than additive, and the ambient noise background i s non Gaussian, or even strongly non Gaussian at low grazing angles. This paper addresses this extremely difficult problem with the aid of two simplifying assumptions, 1 ) the signal modulation i s weak and 2) the departure from Gaussianity is small. In situations where the departure from Gaussianity is large, an approach is suggested for reducing this non Gaussianity. The relevant weak signal detection theory, based on the Likelihood ratio, is reviewed and adapted for use in the analysis. The approach to this problem, similar to that previously used by the author for complex images, is facilitated by approximating the multivariate probability distributions as a composite integral involving underlying processes which are assumed to be Gaussian. This formulation, subject to the approximations in the analysis, permits the derivation of an ideal detection statistic (which determines the form of the optimum receiver) and a SNR which characterizes detection performance in the weak signal limit. Some implications for image processing are discussed and directions for future analysis are suggested