The principle of the Elman nueral network and the features of small infrared target detection are presented in this paper. The Elman neural network is used to estimate the background in an infrared image and then the background noise is eliminated according to the principle of background elimination. Thus, the potential small infrared target is detected. The Neural Network Toolbox in Matlab7.0 is used to do an experiment on programming. The experimental result shows that this method is effective for the detection of small infrared targets.