基于开关型非局部均值滤波的指静脉图像去噪  

Finger Vein Image Denoising Based on Switching Non-local Mean Filtering

作　　者：杨航 沈雷[1] 李凡[1] 吕葛梁 YANG Hang;SHEN Lei;LI Fan;LV Ge-liang(School of Communication Engineering,Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学通信工程学院

出　　处：《软件导刊》2020年第1期262-266,共5页Software Guide

基　　金：杭州电子科技大学研究生科研创新基金项目(CXJJ2018029)

摘　　要：为了解决带有皮裂纹特征的指静脉图像中干扰区域影响静脉特征提取的难题,提出一种基于开关型非局部均值滤波的指静脉图像去噪算法。该算法首先生成静脉图像像素点的Hessian矩阵,然后通过Frangi滤波对起皮干裂区域进行精准检测,最后对受干扰区域进行非局部均值滤波去噪。理论分析与实验结果表明,该算法与现有经典静脉去噪算法相比,充分利用了图像上其它区域的冗余信息,在实现去噪的同时,更好地保留了非受干扰图像区域的静脉细节部分,一定程度上克服了经典去噪算法对整个图像的过平滑问题。在认假率为0时,拒真率相比原结果降低了5.63%。In order to solve the problem that the disturbed area in the finger vein image with the peeling cracking feature affects the vein feature extraction,a finger vein image denoising algorithm based on switching non-local mean filtering is proposed.The algorithm first generates the Hessian matrix of the pixel points of image.And then the peeling and cracking area is accurately detected by Frangi filtering.Finally,and the disturbed region is denoised by non-local mean filtering.Experimental results and theoretical analysis indi⁃cate that compared with the existing classical vein denoising algorithm,the proposed algorithm makes full use of the redundant informa⁃tion of other areas on the image,and better preserves the vein detail of the non-disturbed image area.To a certain extent,it overcomes the problem that the classical denoising algorithm is too smooth for the entire image.In the case of the false acceptance rate of 0,the false rejection rate is reduced by 5.63%compared to the original result.

关 键 词：HESSIAN矩阵 Frangi滤波 非局部均值滤波 图像去噪 指静脉识别 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]