基于多尺度冗余字典稀疏分解的纸病图像背景补偿方法  

Background Compensation Method of Paper Defects Image Base on Sparse Decomposition via Multi-scale Redundant Dictionary

作　　者：周强[1] 杜晞盟 王志强[1] 

机构地区：[1]陕西科技大学电气与信息工程学院,陕西西安710021

出　　处：《中国造纸》2016年第11期45-51,共7页China Pulp & Paper

基　　金：陕西省科技攻关项目(2016GY-005);陕西省科技攻关项目(2011K06-06);陕西省科技统筹创新工程计划项目(2012KTCQ01-19);西安市未央区科技计划项目201304

摘　　要：对于纸病检测中纸张图像背景不均匀以及图像灰度特征不明显等造成纸病测量精度低的问题,建立多尺度冗余字典,采用正交匹配追踪算法(OMP)对纸病图像进行稀疏分解,并根据纸病背景图像和纸病图像不同形态特征,对背景进行补偿,从而增强纸病特征。实验表明,该方法能够有效地重构并补偿纸病背景图像,突出灰度特征较弱的纸病,提高纸病检测的准确性。Aiming at the low measurement accuracy in paper defect detection due to the non-uniform image background and unobvious gray level feature, the paper suggested adopting Orthogonal Matching Pursuit（OMP） to conduct sparse decomposition by establishing a multi-scale redundant dictionary, the image background was compensated based on the background image and the different characteristics of paper defect image, thereby enhancing the paper defect characteristics. Experiment showed that this method could efficiently restructure and compensate the background image, highlight the paper defects with low gray level characteristic, eventually to improve the accuracy of paper defect detection.

关 键 词：纸病检测 稀疏分解 正交匹配追踪算法 多尺度冗余字典 图像背景补偿 

分 类 号：TS77[轻工技术与工程—制浆造纸工程] TP391.41[自动化与计算机技术—计算机应用技术]