Enhanced target tracking through infrared-visible image fusion

Video surveillance is an important tool for force protection and law enforcement, and visible and infrared video cameras are the most common imaging sensors used for these purposes. A feasibility study on fusing concurrent visible and infrared imageries using 13 spatial domain and pyramid-based pixel-level fusion algorithms to improve the tracking performance of an existing video surveillance system was performed. Some of the decomposition methods were designed to increase the contrast, whereas some wavelet methods offered shift invariance. The effects of these fusion algorithms on the detection and tracking performance of the given target tracker were examined and compared. Fusion method based on the ratio of low-pass pyramids was shown to offer a superior detection performance at a relatively low computational cost.