Markov Random Field Models for Non-Quadratic Regularization of Complex SAR Images

This paper presents a comparison between Markovian models for Synthetic Aperture Radar (SAR) image despeckling within the complex domain. The novelty of this paper is enhancement of single look complex SAR images and information extraction. The Gauss-Markov Random Field model, Auto-binomial and Huber-Markov Models are used with the non-quadratic regularization. The experimental results using synthetic generated images and real SAR images showed that the best results were obtained with the Auto-binomial model followed by the Gauss-Markov Random field, and finally the Huber-Markov model, for synthetic generated data and real single look complex SAR images.