基于深度学习的低光彩码图像增强  

Low-light color code image enhancement based on deep learning

作　　者：方路平[1] 翁佩强 周国民[2] FANG Luping;WENG Peiqiang;ZHOU Guomin(College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023,China;Department of Computer and Information Technology,Zhejiang Police College,Hangzhou 310053,China)

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310023 [2]浙江警察学院计算机与信息技术系,浙江杭州310053

出　　处：《浙江工业大学学报》2020年第4期384-391,共8页Journal of Zhejiang University of Technology

基　　金：国家自然科学基金资助项目(U1509219)。

摘　　要：彩码作为一种新型识别码相比于传统二维码具有远距离识别、多识别等优点。但由于彩码图像包含红、绿、蓝三原色色块,导致彩码图像在识别过程中对亮度和对比度有较高的要求。在现实环境中对彩码的识别过程中会有各种复杂的识别场景,因此,针对不同低光条件下的彩码图像,提出了一种基于卷积神经网络的低光彩码图像增强算法。通过卷积神经网络预测低光彩码图像的光照图像,再通过Retinex模型获取增强后的彩码图像。研究结果表明:该方法不仅能对不同低光程度的彩码图像有针对性的增强,也能较好地避免彩码图像色彩的失真。As a new type of identification code,color code has the advantages of long-distance recognition and multiple recognition compared with traditional two-dimensional code.However,since the color code image contains three primary color blocks of red,green,and blue,the color code image has higher requirements for brightness and contrast in the recognition process.There are various complex recognition scenarios in the process of identifying color codes in the real environment.Therefore,aiming at the quality problem of color code images under different low light conditions,an image enhancement algorithm based on convolution neural network is proposed.The light image of low-luminance code image is predicted by convolution neural network,and the enhanced color code image is obtained throughretinex model.The experiment shows that this method can not only enhance the color image with different low light levels,but also avoid the distortion of color image.

关 键 词：图像增强 Retinex模型 卷积神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]