Detection of Dim Infrared Targets by Multi-Direction Prediction of Background

The detection of dim targets in clutter background is an important problem in infrared image processing. The traditional two-dimensional least mean square (TDLMS) filter has been used to detect the dim targets in infrared image sequence recently, but there are some problems with these methods such as a bad adaptability with complex scenes, anisotropy, and a high probability of false alarm (PFA) on the edge of clutter. By analyzing the directional feature of TDLMS filter, we fused the four neighborhood filters and then got a detecting algorithm based on four-direction fusion adaptive two-dimensional least mean square filters (FD-TDLMS) at last. Simulations were conducted on the detecting of dim targets in cloud background and artificial images, and they have shown that this algorithm has a better adaptability with complex background, a lower false alarm probability on the edge of clutter than traditional method when has a good detect probability at the same time.