Abstract: There exist cases when fractal features are used to the target detection and recognition that the fractal dimensions of targets are often close to that of their backgrounds. In these cases, it is very hard to distinguish targets from their backgrounds. The solution put forward in this paper is to calculate the fractal features based on the wavelet transform of the image. The image segmentation in the frequency field through wavelet transform enriches the illustration of the image, and the fractal dimensions calculated from the detail images of different frequency segments can be made to a united characteristic vector which is helpful to distinguish man-made targets from complicated background rapidly and accurately. The experimental results indicated that man-made targets and natural background could be effectively distinguished by this method. But it should be mentioned that the mass calculation needed might limit the application of this method.