Abstract: For detecting the moving dim-small targets in infrared image sequences, the paper studies the motion characteristics, the probability distribution characteristics, and the energy distribution characteristics between targets and noises. Firstly, it uses high-order cumulant based on M continuous images to enhance the energy of moving dim targets. Then the hypothesis test is used to separate the targets and background. Finally, the multi-frame correlation detection is made through searching the moving energy center of image sequences. The simulation results show that the proposed algorithm is effective for infrared dim-small targets detection.