Abstract: Aiming at the problem of the high false alarm rate of detecting dim and small targets in the infrared image with low signal-to-noise ratio and complex background, a saliency map based dynamic programming track-before-detect algorithm is proposed. An improved difference of local region operator is used to extract the saliency map in this algorithm. A novel search strategy is designed according to the attention transfer mechanism. In addition, the search range is updated in real time by utilizing the speed of the target. The algo- rithm processes the multi-frame continuous saliency maps to realize infrared dim and small target detection. Experiment results indicate that the algorithm can reduce false alarm rate and improve detection performance in less time.