Statistical segmentation of radar images

Considers the problem of unsupervised segmentation of radar images. These radar images, representative of the azimuth-distance space, have been segmented at differences resolutions. The coarsest resolution radar images, coming from a maximum computation on few APC cells in distance by azimuthal sectors (Anti Clutter Processor: APC pixels (or cells) are formed by a temporal average maximum computation on a window of few radar cells extended in azimuth and distance). These types of images characterize radar clutters (clouds, sea, chaff, snow, rain, ground, angel) and permit one to analyse their spatial resolution. Statistical segmentation provides a useful tool for detection of targets in clutters, more efficient than thresholding in case where targets own an azimuthal or distance spreading, because algorithms take into account spatial correlations to extract different statistical populated area of pixels in radar image