Abstract: Most median filtering algorithm based on the process of sorting theory contains a large number of comparison and data shifting, slows down the speed of algorithm. In considering the filtering data word length and specific application based on the characteristics of the processor, a kind of fast median filtering algorithm is put forward. By taking advantage of the correlationship of previously sorted data, the initial position of the latest acquisited data is quickly determined. By taking advantage of direct dichotomy, the final position of the latest acquisited data is quickly determined, and the number of comparison is declined to n(N=2ⁿ-1, here N is for the length of filtering window). By designing mapping arrays, the number of acquisited data shifting is declined to zero. Through simulating Aduc847 Micro-converter on the Keil C software platform, experimental results show that this algorithm can be used in the system of 1 kHz real-time data acquisition.