Image filter is the method for image pre-processing in many image processing algorithms. Since, median filter not only can smooth image noise, but also can hold the image edge, it becomes the most used method in image filter. Median filter needs sorting operation, its time complex is very high. When filter window becomes larger, median filter will need more time, it cannot be used in large image real-time processing. Based on the median filter research, a faster median filter algorithm which is based median division is proposed. The proposed algorithm can reduce filter time effectively, especially in large scale image, when filter window is large, it becomes more remarkable. Compared with the traditional median filter algorithm, the proposed algorithm filter result can hold the preference as same as the traditional median filter algorithm, and it can be used on gray and color image. The experiment shows that the proposed median filter algorithm not only speeds up image filter speed, but also improves the real-time of image processing.