Abstract

The electromagnetic interaction between microwaves and sea surface makes the sea clutter in SAR images with specific statistical properties. However, speckle noise, bathymetry and local winds fields, especially in coastal areas, may complicate the SAR image interpretation. In fact, they deeply affect the homogeneity of the clutter statistics, thus complicating the detection of maritime targets. CFAR processing schemes can be used to overcome this problem. Such methods set the threshold adaptively based on local information of the clutter noise power. The threshold in a CFAR detector is set on a cell by cell basis using estimated clutter power by processing a group of reference cells surrounding the CUT (cell under test). In a homogeneous background when the reference cells contain independent and identically distributed (i.i.d.) observations (governed by exponential distribution) the CA-CFAR processor is the optimum CFAR processor. The CA-CFAR (cell average-CFAR) procedure uses the maximum likelihood estimate of the noise power to adaptively set the threshold under the assumption of i.i.d noise samples. However, the CA-CFAR performances are significantly affected when the assumption of clutter homogeneity is violated. Moreover, conventional CFAR-based algorithms could be very time consuming because they inspect all the pixels of a SAR image.