基于强度和梯度先验的L_0正则化模糊QR码识别  

Recognition of Fuzzy L_0-Regularized QR Code Based on Intensity and Gradient Priori

作　　者：杜菲[1] 曾台英[1] 

机构地区：[1]上海理工大学,上海200093

出　　处：《包装工程》2017年第3期150-154,共5页Packaging Engineering

基　　金：中科院上海光学精密机械研究所重点实验室基金(3A13309013)

摘　　要：目的研究因机械抖动,拍摄器材与图像存在一定距离或相对运动而产生运动模糊、散焦模糊等情况下的模糊QR码图像识别。方法采用基于强度和梯度先验的L_0正则化方法对模糊QR图像进行去模糊。优化模糊核尺寸的人为预估问题,提高程序效率。对1至15类常用QR码图像进行模糊仿真,再通过盲提取获得模糊核,用峰值信噪比PSNR值衡量该方法在QR码图像去模糊的复原精度。结果PSNR值随着QR码图像复杂度的增加而相对减少,但因QR码存在一定的容错率,在PSNR值为13以上且噪声、振铃小的情况下就能够被识别。文中算法相较于其他算法在型号较高的模糊QR码恢复方面识别率更高。结论基于强度和梯度先验的L0正则化方法对模糊QR码的恢复效果显著,且不是只针对某一类模糊QR码图像,对于多种类型的模糊QR码图像恢复都能有很好的效果。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 L_0-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 L_0-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.

关 键 词：QR码 正则化 去模糊 识别 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]