Quick Response (QR) barcode is a quite common two dimensional barcode. Image preprocessing is a key step for the automatic recognition of QR barcode decoder. In order to lower the used threshold, a practical image preprocessing method was proposed for QR barcode recognition. It could increase the speed of recognition by this decoder so as to embed this algorithm into mobile terminals. It didn't utilize the traditional methods such as edge detection and line detection, but only the encoding characteristic of QR to locate, thus the influence of geometric distortion and background noise was reduced. Moreover, in order to improve the recognition rate, it used the rectification images to adaptively sample the QR code in terms of regions. Experiment results demonstrate that the proposed approach can overcome the influence of noise, inhomogeneous light and geometric distortion, and meet the requirements of real-time decoding.