基于指导滤波的内侧指横纹识别方法  

Method for Inner-knuckle-print Verification Based on Guided Image Filtering

作　　者：刘明[1] 闫军[1] 李哲[1] 

机构地区：[1]河北大学电子信息工程学院,河北保定071002

出　　处：《电视技术》2014年第11期202-205,233,共5页Video Engineering

基　　金：国家自然科学基金项目(60903089;61073121);河北省自然科学基金项目(F2009000215);河北省科技支撑计划项目(072135188);河北省教育厅科研计划项目(2008312);北京市优秀博士学位论文指导教师科技项目(YB20081000401)

摘　　要：为了使指导滤波可以更有效地去除噪声进而提高匹配效率,提出了一种基于指导滤波的内侧指横纹识别方法。首先,在ROI提取阶段,通过提取ROI模板和利用模板在原始图像上所处位置的先验知识来指导其他原始图像进行准确的ROI提取,并对得到的ROI和模板进行互相指导滤波去除噪声。其次,在线特征提取阶段,利用Gabor滤波与求偏导相结合的方法进行线特征提取。最后,在特征匹配阶段,利用快速傅里叶变换的方法,得到相似度后通过对其选取合适的阈值来得到较好的实验效果。评估系统建立在包含100个人的2 000幅右手手部图像的数据库上,等错误率为2.32%,证明该算法可以进行指横纹特征识别并具有较高的准确率。In order to make the guided image filtering can remove noise effectively and improve the matching rate, an approach for inner-knuckle-print verification is proposed based on guided image filtering. Firstly, in the stage of ROI extraction, by extracting ROI template and using the location on the original image to guide other images to extract ROI accurately, then adopt mutual guided filter between the ROI and the template. Secondly, in the stage of line detection, the combination of Gabor filter and derivative line detection is adopted. Finally, the binary line images are matched by using Fast Fourier transform, after get similarity by selecting appropriate threshold to get better effect. The experiments on a finger image database which includes 2 000 images from 100 different individuals. The equal error rate reaches 2.32%. The experiment result shows good performance of the proposed approach.

关 键 词：生物特征 内侧指横纹 指导滤波 线特征提取 

分 类 号：TN911.73[电子电信—通信与信息系统]