应用超分辨率重建算法的图像匹配算法  

Image matching algorithm using super-resolution reconstruction

作　　者：武玉坤[1,2] 陈沅涛 WU Yu-kun;CHEN Yuan-tao(School of Management and Information,Zhejiang Post and Telecommunication College,Shaoxing 312016,China;College of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310023,China;School of Computer and Communication Engineering,Changsha University of Science and Technology,Changsha 410114,China)

机构地区：[1]浙江邮电职业技术学院管理与信息学院,浙江绍兴312016 [2]浙江工业大学计算机科学与技术学院,浙江杭州310023 [3]长沙理工大学计算机与通信工程学院,湖南长沙410114

出　　处：《计算机工程与设计》2019年第12期3569-3574,共6页Computer Engineering and Design

基　　金：国家自然科学基金青年基金项目(61602060);浙江省2016年度高校访问工程师“校企合作”基金项目(FG2016192)

摘　　要：图像场景中提取特征数量较少,难以满足三维重建对特征匹配的要求,为此提出基于超分辨率重建的特征提取算法。采用回归学习对低分辨率图像进行重建,对重建结果采用图像先验知识优化,获取较高分辨率图像,对高分辨率图像进行特征提取。实验结果表明,相较重建前图像平均提高6倍以上特征点数量,提高5倍以上匹配点数量,该算法能够提高提取特征点的数目,满足三维重建等需要较多特征点的情况。Generally,the amount of feature points obtained from images is so small that it can’t satisfy the demands of dense 3-D reconstruction and some other situations.To solve the problem,the feature extract algorithm based on super-resolution reconstruction procedure was proposed.Low-resolution images were reconstructed using regression algorithm and the results were optimized based on the priori knowledge of natural images.The super-resolution images were obtained and feature extraction was implemented.Experimental results show that in contrast with the low-resolution images,the feature points amount can be increased by six times and the matching points by five times.The proposed algorithm can increase a large number of feature points.It can meet the demands of dense 3-D reconstruction and some other situations.

关 键 词：图像匹配 超分辨率重建 回归学习 特征点数量 图像先验知识 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]