Analysis of multi/hypersspectral images - progress and results

This report describes three different image processing methods with three different purposes. All three are necessary for the design of sensor systems to enable the sought-for capability to, during reconnaissance of an area, detect anomalies, changes, and specifc spectral signatures (materials). The first method is for registration, ie to find the spatial transformation between two images depicting (approximately) the same area. Correct registration is needed for change detection, mosaicking, image fusion, and super resolution. The described method exploits known methods and adapts them to military needs. The second method is for atmopsheric correction in order to be able to study the spectral signatures of targets in the reconnaissance area. The third method is an information theoretic measure for sensor per- formance assessment and is intended to give a useful measure independent of detection methods.