The detection of unusual events in remote sensed imagery

Describes a novel technique for the detection of unusual events in remote sensed imagery when the images contain a high level of background clutter. The technique combines conventional image processing algorithms with statistical methods to derive simple models of image backgrounds. These fractal-based models have been shown experimentally to give a good fit to statistics derived from remote sensed imagery of structured natural backgrounds such as mountain ranges and the sea surface. Unusual events, such as man-made objects show statistics that differ markedly from the background models and can be detected by finding differences between measured image statistics and the assumed background model.<
>