Multi-resolution algorithms for clutter rejection

The detection of both point and extended targets in strongly cluttered environments, using spatial information, is presented. On the basis of an adaptive statistical model of the background clutter, the existence of a possible target region is detected by the triggering of a low-probability threshold, indicating an unlikely event when interpreted in the context of the cluttered background. This strategy is implemented in a proposed pre-processor whose purpose is to designate limited potential threat regions within the overall image, to which detectors incorporating prior knowledge of threat objects would subsequently be directed. By analogy with biological sensory systems, the pre-processor has a pre-attentive role, designed to achieve extensive preliminary data reduction in the form of clutter rejection