基于信息熵的二维局部二值模式静脉识别
Information entropy-based vein recognition for two-dimensional local binary patterns
张云飞 1李江美 1陈熙1
作者信息
- 1. 贵州师范大学大数据与计算机科学学院,贵阳 550025
- 折叠
摘要
基于现有LBP算法及其变体无法提取图像高维特征的问题,提出一种基于信息熵的二维局部二值模式识别算法.此方法首先利用统一局部二值模式(ULBP)对图像进行低维特征的提取,随后将图像信息熵与统一局部二值模式图谱进行结合获取熵值加权的统一局部二值模式图谱(EULBP),并利用滑动窗口实现对局部区域内模式间共现特征信息的统计,以其结果作为图像特征表达.并以直方图交叉距离为基础构建模式分类器,验证其识别性能.实验结果表明,在SDUMLA-HMT数据集以及马来西亚理工大学指静脉数据集(FV-USM)中,提出的算法能取得99.94%和98.84%的平均识别率.
Abstract
Based on the problem that the existing LBP algorithm and its variants are unable to extract high-dimensional fea-tures from images,a two-dimensional local binary pattern recognition algorithm based on information entropy is proposed.This method first extracts the low-dimensional features of the image using the unified local binary pattern(ULBP),then combines the im-age information entropy with the unified local binary pattern atlas to obtain the entropy-weighted unified local binary pattern atlas(EULBP),and realizes the statistics of the co-occurring feature information among the patterns in the local area using the sliding window,and uses the result as the image feature expression.And the pattern classifier constructed on the basis of histogram cross distance is used to verify its recognition performance.The experimental results show that the proposed algorithm can achieve an av-erage recognition rate of 99.94%and 98.84%in both the SDUMLA-HMT dataset as well as the Universiti Teknologi Malaysia fin-ger vein dataset(FV-USM).
关键词
二维共现局部二值模式/信息熵/旋转不变/方向特征Key words
two-dimensional co-occurrence of local binary patterns/information entropy/rotational invariance/directional features引用本文复制引用
基金项目
国家自然科学基金(61762022)
湖南省教育科学规划课题(十四五)(2023)(XJK23BJC011)
出版年
2024
NETL
NSTL
万方数据