Abstract: Infrared face recognition, being light- independent, and not vulnerable to facial skin, expressions and posture, can avoid or limit the drawbacks of face recognition in visible light. In this paper, a fast infrared face recognition system based on local features was proposed. Firstly, the original IR images captured by an IR camera thermoVision A40 were 320240 pixels. The sensitivity was as high as 0.08 ℃, the face image was normalized to size of 6080 by preprocessing and face detection. Secondly, based on the symmetry of face, the whole infrared face image was divided into small sub-blocks. To make full use of the local features of infrared face image, a local binary pattern (LBP) was chosen to get the composition of micropatterns of sub-blocks. Finally, Kruskal-Wallis (KW) feature selection method was proposed to improve the effectiveness of discriminant feature extraction. The experiment results illustrate that the system recognition rate can reach 98.6%, outperform the traditional methods based on Principal Component Analysis(PCA) and Linear Discriminant Analysis (LDA). Furthermore, the speed of proposed system is much faster than the traditional methods and can be used in real-time face recognition system.