Infrared images have the characteristics of low contrast,big noise,wide dynamic range and poor visual effect,etc. In order to obtain a good visual effect and improve the target detection and recognition capabilities,Retinex enhancement algorithm based on human vision was studied. Combined with imaging mechanism of infrared image,Retinex method that is suitable for the visible light image is used for infrared image enhancement,and the ideal visual effect has been gotten through this method. On this basis,an adaptive multi-scale Retinex (AMSR) method is proposed. The infrared images is enhanced by this algorithm and the objective evaluation factors are calculated,and its enhancement effect compares with those of several traditional image enhancement algorithms. The experimental results show that the AMSR algorithm has the applicability and superiority for infrared images. This study is significant to the infrared target detection and recognition.