QR code images fast blind deblurring based on specific symbol prior and L2 norm minimization

Recently QR code has been widely used and has made great convenience to our daily life. However, the degradation of QR code images will make them difficult to be recognized and decoded. In the meantime, we find there are some problems when the existing excellent deblurring algorithms are applied to the QR code images. First, the restored over-sharpened edges may destroy the ratio of specific symbol in images, which will lead to the failure of decoding. Second, these methods are time-consuming and impractical to be directly used. To solve these problems, a model based on $L_{2}$ norm is proposed to avoid over-sharpened edges, and then the gray level and gradient features of QR code are utilized to optimize it. Besides, based on the specific symbol prior, we introduce an evaluation mechanism to achieve fast deblurring. The experimental results show that our method has better performance than other state-of-the-art algorithms.