HIDDEN MARKOVMODEL-BASED MULTI-MODAL IMAGE FUSIONWITH EFFICIENT TRAINING