基于GEP多标记学习的图像超分辨率复原算法  

Image super-resolution restoration algorithm based on GEP multi-label learning

作　　者：汤嘉立 柳益君[1] 杜卓明 Tang Jiali;Liu Yijun;Du Zhuoming(College of Computer Enginering,Jiangsu University of Technology,Changzhou Jiangsu 213001,China;School of Mechanical Engineering,Jiangsu University,Zhenjiang Jiangsu 212013,China;College of Science of Mathematics,Nanjing Normal University,Nanjing 210023,China)

机构地区：[1]江苏理工学院计算机工程学院,江苏常州213001 [2]江苏大学机械工程学院,江苏镇江212013 [3]南京师范大学数学科学学院,南京210023

出　　处：《计算机应用研究》2018年第6期1849-1852,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61402206);中国博士后科学基金资助项目(2016M601845);国家住房城乡建设部研究开发项目(2016-K8-028);江苏理工学院应用基础研究计划基金资助项目(KYY14005)

摘　　要：针对传统基于范例学习超分辨率复原算法的样本块误匹配和结果不稳定等不足,提出一种基于基因表达式编程多标记学习的超分辨率复原算法,筛选出与目标图像相关性高的样本子库,在多标记框架下进行样本预分类。该算法根据图像的多重特征筛选出其相关图像类别,离线建立分类模型,提高了图像质量和计算速度。实验结果表明,该算法稳定性强、鲁棒性好,缩小了低分辨率图像块的匹配范围,提高了超分辨率复原的效果和效率。Tranditional example-based restoration algorithms easily lead to sample mis-matching and unstable recovered results because of the diversity of image features. For such problems,this paper proposed a super-resolution restoration algorithm based on GEP multi-label learning. The restoration algorithm selected the subset of sample library which was highly related to the object image and pre-classifies the image samples under the multi-label framework. It chose the related category according to the image multi-figures and built off-line disaggregated models,which improved the image quality and running speed. The experimental results show that the proposed method is robust and stable. Specifically,the algorithm further reduces the matching range of low resolution image blocks and promotes the restoration effectiveness and efficiency.

关 键 词：超分辨率复原 基因表达式编程 支持向量机 样本学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]