A Novel Approach to Robust Blind Classi cation of Remote Sensing Imagery

In this paper we propose a novel method for the robust classi cation of blurred and noisy images that incorporates ideas from data fusion. The technique is applicable to blind situations in which the exact blurring function is unknown. The approach treats di erently deblurred versions of the same image as distinct correlated sensor readings of the same scene. The images are fused during the classi cation process to provide a more reliable result. We show analytically that the various restorations can be treated as images acquired from di erent but correlated sensor readings. Experimental results demonstrate the potential of the method for robust classi cation of imagery.