Pedestrian tracking in infrared image sequences using wavelet entropy features

In this paper, we deal with the problem of accurate pedestrian tracking in infrared image sequences. We introduce a wavelet entropy features based representation for pedestrians in infrared imagery and study its application into pedestrian tracking using sequential Monte Carlo methods. Firstly, the regions of interest's (ROI) representation is constructed using wavelet entropy which hurdle the disadvantage of insufficient information when only intensity feature is considered. Then the aforementioned pedestrian representation model embedded in the particle filter framework is updated and sample distributions are propagated over time. Experimental results using infrared image sequences are reported to demonstrate the encouraging performance of our algorithms.