A FPGA-based real-time infrared image acquisition and pre-processing system is designed. Beginning from the overall design, the design principle of the system and each component of the hardware circuitry are presented. Since the image processing system in the current target range equipment may destroy the detail in an image while it reduces the impulse noise in the same image effectively, a cross-window-based method for implementing fast median filter algorithm in FPGA in real time is proposed in combination with the advantages of FPGA in the aspects of parallel structure and pipelining. The high-level algorithm in a target range is replaced with this method and much precious processing time is saved. In this system, several technologies such as modular design, assembly line and ping-pong operation are used. The experimental result shows that the method has a better noise suppression ability and can improve the instantaneity of the entire system. So, it is of great value in practical applications.