For solving the pre-processing problem of dim small target detection in infrared images, a characteristic-selecting filtering (CSF) method is proposed. The method applied anistropic diffusion to images’ spatial domain creatively. Through analyzing the properties of various kinds of characteristic regions, the method improved the coefficient distribution function to suppress clutter, smooth edge and enhance target. The filtering characteristic analysis and experimental results were given, which were compared with those by some other methods. The results show that this approach can provide good performance of suppressing clutter and enhancing target signal, and its structure is simple to be implemented in real-time system.