A Bayesian approach for SAR images segmentation and changes detection

In the context of image processing and classification, an important problem is the development of accurate models for Synthetic Aperture Radar (SAR) image segmentation. In this paper we propose a highly efficient unsupervised algorithm for image segmentation and changes detection, based on the Generalized Gaussian mixture model. Our work is motivated by the fact that SAR images are highly corrupted by speckle noise, and contain non-gaussian characteristics impossible to model using rigid distributions. Generalized Gaussian mixture models are robust in the presence of noise and outliers, more flexible to adapt the shape of data, and less sensible for over-fitting the number of classes compared to Gaussian mixture.