Probability distribution mixture model for detection of targets in high-resolution SAR images

In this paper, the detection of close targets in heterogeneous clutter in high-resolution SAR images is investigated. We adopt a probability distribution mixture model where each pixel intensity image is characterised by two probability density functions: one related to the targets and one related to the background clutter. A specific detection threshold, based on the estimates of the mixture parameters, is used. The statistical characterisation of SAR images modeling is a key issue for detection. The clutter is modelled using the K distribution that is a flexible tool over non-homogenous areas. We show that our method is able to detect close targets at constant false alarm ratio without making any assumptions about their size and their spatial configuration.