Multilogit Prior-Based Gamma Mixture Model for Segmentation of SAR Images

Synthetic Aperture Radar (SAR) has the capability of working in all weather conditions during day and night that make it attractive to be used for target detection and recognition purposes. However, it has the problem of speckling that is structured as multiplicative noise which makes the SAR data a complex image. The algorithms need to be sufficiently robust to speckle noise for the achievement of reliable segmentation from such complex images. In this letter, the first contribution is the development of a robust multilogit spatial interactive model as a categorical distribution. The categorical property of this approach makes it ideally suited to be used as a pixel-based prior to any finite mixture model. Second, multilogit spatial interactive gamma mixture model is developed which is based on this prior. Experimental results with synthetic and real images indicate that the proposed mixture model is highly effective in segmenting SAR images.