When performing inter-satellite laser communication, the accuracy of spot position detection will directly affect the capture, tracking and aiming process. However, due to interference or defects in the response of its own device during the acquisition and transmission process, blur and noise appear, so filtering preprocessing should be performed before spot positioning. The classical median filtering algorithm of 3×3 window had a good filtering effect on noise, but the processing time was too long. Therefore, an asymmetric fast median filtering algorithm based on 3×4 window was proposed. First, the 6 intermediate pixel values reused in the 2 adjacent windows of the classical method were sorted, and then compared with the median values in the left-most column and right-most column of the window respectively. The obtained median values replaced the 2 pixels in the original signal center position. By visual observation, the denoising effects of the 2 filtering algorithms were equivalent. For the spot location error caused by non-improved median filtering of 3×4 window, the unified standard comparison could be used to compensate, and the error was within an acceptable range. The simulation results showed that the improved median filtering algorithm shortened the response time. Therefore, the improved asymmetric median filtering algorithm enhanced the real-time performance of the algorithm and was more suitable for the design and implementation of FPGA-based platforms.