Based on the study of directional derivative characteristics of Facet model， a detection method for infrared weak targets is proposed. This detection method uses local relative extreme differential value at the multi-directional lines to calculate the significance of small target, which is very effective for fast extraction of targets under complexbackground. This method is based on the theory of single frame detection of weak target. Firstly, the Facet wizard number characteristics of the original image are calculated. Secondly, in the local part of the Facet wizard number feature graph, the relative extreme difference contrast is calculated along the current direction. Then the relative extreme difference contrast in each direction is fused to obtain the final significant image. Finally, the target is extracted with appropriate threshold segmentation for the final significant image. The experiment result shows that the proposed algorithm has high signal-clutter gain and background inhibitor for complex infrared weak target images. In addition, the computational complexity of the algorithm is low and can be calculated by using two-dimensional convolution acceleration. As a good real-time algorithm, it is beneficial to the engineering implementation of various processor platforms.