Target tracking using the anisotropic diffusion pyramid

Forward-looking infrared (FLIR) systems are used to identify targets based on their features and emissions. With the new generation of FLIR systems that can produce sixty or more high resolution frames per second, multiscale algorithms that can operate accurately and efficiently in a variety of battlefield situations are paramount. The popular Gaussian pyramid tends to distort and even destroy object edge information with increasing iterations, leading to incorrect target identification. We propose a new approach to infrared searching and tracking based on the anisotropic diffusion pyramid (ADP). In contrast to the Gaussian pyramid, the ADP preserves edge information while increasing search speed.