Bayesian Reconstruction And Texture Segmentation Of Sar Images
Marc Walessa, Mihai P. Datcu · 1998
An approach to unsupervised segmentation of SAR images and texture modelling is presented. This approach is based upon a Bayesian estimation in order to obtain the maxi- mum a posteriori segmentation of an image. The segmentation is performed by using prior information describing the features according to which an image is to be segmented and, where mathematically tractable, takes into account the likelihood function of the SAR specific speckle statistics. The solution of this optimization problem is obtained by a relaxation method combined with a region growing approach.