The work aims to study the recognition of fuzzy QR code images caused by the motion blur and defocus blur of mechanical vibration, and the certain distance or relative motion between photographic device and the image. The L0-regularized method based on intensity and gradient priori was used to deblur the fuzzy QR code images. The problem of artificial estimation of fuzzy kernel size was optimized and the program efficiency was improved. Blurring simulation for 1 to 15 kinds of common QR code images was carried out, and then blurring kernel was obtained by blind extraction. PSNR value was used to measure the restoration precision of the method in deblurring QR code images. The PSNR value was relatively decreased with the increase of the complexity of the QR code images; however, because the QR code had a certain fault tolerance rate, it could be recognized when the PSNR value was above 13 and the ringing & noise were small. Compared with other algorithms, this algorithm had a higher recognition rate in restoring the fuzzy QR codes of higher model. The L0-regularized method based on intensity and gradient priori can restore the fuzzy QR codes remarkably, not just for certain type, but for a variety of fuzzy QR code images.