Abstract: Infrared video in airport surveillance scenario with a large depth of filed consists of smooth background and foreground areas with less textures, compared with traditional video. Therefore, the temporal and spatial correlations of Infrared video is larger. Although commonly used video codecs can be utilized to compress Infrared video, only a little local information is used to reduce the redundancy with high computation cost. In view of this, this paper considers the whole Infrared video as a tensor from global angle and then use tensor decomposition approach to reduce the data volume. The resulting factor matrix is then quantized. Moreover, this paper analyzed the compression performance of CP decomposition and Tucker decomposition. Experimental results show that, compared with HEVC, Tucker decomposition can achieve about 77.7% BD-rate savings while the video contains lots of smooth areas. If not, CP decomposition and Tucker decomposition would BD-PSNR lose about 10 dB. However, the encoding time of CP decomposition and Tucker decomposition is just about 0.57% and 2.25% compared with HEVC, respectively.