Influenced by many factors, the obtained Infrared images usually have a low signal-to-noise ratio (SNR) and contrast, which brings a certain difficulty to target extraction. Based on analyzing the direction gradient of the target, with morphology algorithm, a new detection algorithm of small Infrared targets is introduced. Firstly, four reference points at the direction of four neighbor domain of the tested point are chosen. The potential targets are determined according to multi-level gradient feature between the tested point and the reference point at the direction. And then, the adjustable feature of structure element is used to choose appropriate structure element. Finally, through morphology processing, noise points are eliminated and the target is determined. The algorithm has simple calculation and in no need of predicting background. Experimental results show that small targets can be detected effectively in low signal-to-noise ratio (SNR) images.