We show how nonlinear filters tuned to the spatial and temporal frequency of a moving target may be used to selectively detect and track certain specified moving objects in an image sequence, while ignoring other moving objects as well as background movement. Further, we describe an algorithm for adaptation of the filter parameters to maximize the strength of the filter output as the target changes in speed or size. We characterize the performance of these tunable filters in detection and tracking of a moving target in infrared imagery, first for artificially-generated scenes, and then for real infrared imagery.
Anusha Muthu-Natarajan, "Adaptive Spatio-Temporal Filters for Infrared Target Detection," MS thesis (Advisor: Charles M. Higgins), Department of Electrical and Computer Engineering, The University of Arizona, November 2005.