基于多特征融合的手指静脉图像检索方法  

Finger vein image retrieval based on multi-feature fusion

作　　者：刘行 杨璐 郝凡昌 LIU Xing;YANG Lu;HAO Fanchang(School of Computer Science and Technology,Shandong Jianzhu University,Jinan 250101,Shandong,China)

机构地区：[1]山东建筑大学计算机科学与技术学院,山东济南250101

出　　处：《山东大学学报（工学版）》2023年第2期118-126,共9页Journal of Shandong University（Engineering Science）

基　　金：国家自然科学基金资助项目(62076151);山东省重点研发计划公益类资助项目(2019GGX101068)。

摘　　要：为进一步提高手指静脉图像检索的精度,提出一种基于多特征融合的手指静脉图像检索方法。从手指静脉图像中分别提取纹理特征、方向特征、静脉主干特征,使用改进的多基元直方图方法计算纹理直方图和方向直方图,将两种直方图串联作为检索特征。考虑到手指静脉血管的空间分布特性,对图像进行分块处理,基于图像块提取检索特征。该方法在两个公开手指静脉数据库上的最优检索精度分别为99.16%、99.15%。试验结果表明,该方法能够充分利用多种不同特征,有效提升手指静脉图像检索性能。In order to improve the accuracy of finger vein image retrieval,a finger vein image retrieval method based on multifeature fusion was proposed.In the proposed method,vein backbone,texture feature and orientation map were extracted from finger vein image,and then texture histogram and orientation histogram were defined by an improved multitexton histogram method.Two kinds of histograms were concatenated for finger vein image retrieval.Moreover,for exploring the spatial distribution of finger vein,one image was partitioned into multiple blocks,and multifeature fusion was performed in blockwise.The proposed method was tested on two public finger vein databases,and the retrieval accuracies were 99.16%and 99.15%respectively.The experimental results showed that the proposed method powerfully fused multiple features and improved the retrieval performance.

关 键 词：手指静脉图像检索 改进的多基元直方图 多特征融合 局部二值模式 静脉主干特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]