I. Introduction
Statistical modeling of clutter amplitude or intensity distributions is of great importance in radar signal processing search community [1]. For maritime surveillance radars, sea clutter of quite low resolution can be modeled for instance by the exponential distribution in the intensity domain. However, this model is restricted by some sea surface conditions and radar system parameters. Nevertheless, this model can not characterize reverberation data of higher spatial resolution at low grazing angles and high wind speed. For this reason, a variety of non-exponential probability density functions (pdf) are proposed to model radar sea clutter with heavy tails or ‘spikiness’, including gamma, log-normal, Weibull distributions and five types of compound-Gaussian (CG) distribution [2]. The latter is obtained by the combination of two different pdfs representing the speckle and the texture components. The speckle has an exponential distribution with a power parameter. The latter is called the texture component and is a random variable which can follow gamma, inverse gamma, inverse Gaussian or log-normal distribution [3]. In fact, each texture distribution has a particular representation of sea echoes for a high resolution case. Many non-coherent/coherent CFAR target detection schemes are constructed under these models [4].