基于不变尺度特征变换和有界失真映射的受损文物图像匹配方法

(1.西北大学信息科学与技术学院,陕西 西安 710127; 2.巢湖学院信息工程学院,安徽 巢湖 238000; 3.巢湖学院网络与分布式系统研究所,安徽 巢湖 238000)

保扭曲映射; 三角网格; 复数; 雅可比矩阵; 图像匹配

Matching method of damaged cultural relic images based on scale invariant feature transform and bounded distortion mapping
ZHANG Yong1,2,GENG Guohua1*

(1.School of Information Science and Technology,Northwest University,Xi'an 710127,China; 2.School of Information Engineering,Chaohu University,Chaohu 238000,China; 3.Institute of Networks and Distributed Systems,Chaohu University,Chaohu 238000,China)

bounded distortion mapping; triangular mesh; plural; Jacobian matrix; image matching

DOI: 10.6043/j.issn.0438-0479.2018015

备注

针对受损文物图像这一特殊数据源,提出一种结合不变尺度特征变换(SIFT)和保扭曲映射的图像匹配(SBD)方法.该方法基于SIFT特征建立特征三角网格映射,对网格映射添加保扭曲约束,构建非凸的约束空间,并将属于L0范数问题的目标函数使用代理函数近似.实验结果表明SBD方法可以优化并减少错误匹配,对于解决破损文物图像匹配问题的研究具有较好效果.

An image matching method(SBD)combining scale invariant feature transform(SIFT)and bounded distortion mapping is proposed for the special data source of damaged cultural relic images.The feature triangular mesh mapping is established based on SIFT.In addition,the bounded distortion mapping constraint is added in the method,and the non-convex constraint space is constructed.Furthermore,a proxy function is used to approximate the target function,which belongs to the category of L0 norm.As a result,mismatch pairs can be optimized and reduced,as well as become effective for solving problem of damaged cultural relic images matching.