We propose an unsupervised classification algorithm for high resolution Synthetic Aperture Radar (SAR) images based on Classification Expectation-Maximization (CEM) algorithm. We combine the CEM algorithm with the hierarchical agglomeration strategy and a model order selection criterion called Integrated Completed Likelihood (ICL) to get rid of the initialization and the model order selection problems of the EM algorithm. We exploit a mixture of Nakagami densities for SAR amplitudes and a Multinomial Logistic (MnL) latent model for class labels to obtain spatially smooth class segments. We test our algorithm on TerraSAR-X data.
Koray Kayabol was born in Turkey. He received the B.Sc., M.Sc. and Ph.D. degrees in Electrical & Electronics engineering from Istanbul University, Istanbul, Turkey in 1997, 2002 and 2008, respectively. He was a research assistant in Electrical & Electronics Eng. Dept. between 2001 and 2008. From 2008 to 2010, he was a ICTP Postdoctoral Fellow in the ISTI-CNR, Pisa, Italy. Between September 2010 and July 2011, he was an ERCIM Postdoctoral Fellow in the Ariana at INRIA, Sophia Antipolis, France. Since August 2011, he has been pursuing his ERCIM Postdoctoral Fellowship in the Probability and Stochastic Networks Group at CWI, Amsterdam, Netherlands. His research interests include Bayesian image processing, statistical image models, astrophysical image analysis, image classfication/segmentation and blind source separation.