Infrared image of large field-of-view (FOV) has high resolution, so it is very difficult to process the image in real time by pixel. In order to solve this problem, a method of detection from coarse to fine is put forward to detect the small flying target in a special low-altitude background. We divide image into pieces and compute the entropy, so that the entropy′s segmentation can be performed, complete the background classification and acquire the target area, it′s coarse detection. Then according to top-hat filtering, the dim and small target is detected. The experimental result shows that this method can detect the dim and small target in large FOV more efficiently and accomplish real-time detection.