Abstract: Salt and pepper noise often appears in the acquisition of QR code image. In order to reduce salt and pepper noise as well as maintain image details, median filtering is fequently used in two-dimensional code image pre-processing. However, the typical median filtering algorithm requires a lot of time. Several median filtering methods were applied to the QR code image preprocessing, then the results were compared and QR code image median filtering rules were obtained, the new median filtering method was proposed, which used the window coefficients. Binarization is the first step, then the binary image is processed by median filtering based on filtering window coefficients, and a group of data is obtained through the experimental comparison. The experimental results show the decoding efficiency and accuracy are greatly improved when this coefficients' median fitering is adopted in the QR code image preprocessing.