摘要
首先,采用先行后列的方法对归一化虹膜图像进行经验模态分解,得到不同尺度的固有模态分量;找出有利于识别的分量,将其进行二值化处理生成特征图像;然后对特征图像进行水平和垂直移位匹配,得到海明(Hamming)距离匹配向量,计算匹配向量的改进标准差,以此标准差进行虹膜识别。最后分别对CASIA1、CASIA2、CASIA3-interval、MMU1库进行了识别,结果表明:该方法能够有效地提取图像的二值特征,具有速度快、识别率高等优点。
An iris recognition method based on improved empirical mode decomposition is proposed. First, the normalized iris image is decomposed based by row and then column to generate the different layer intrinsic mode components of the image. Second, the feature image is obtained by binarizing the components useful for the iris recognition. Third, the Hamming distance matching vector is obtained by horizontal and vertical shift match. Finally, the improved standard deviation of the matching vector is calculated, which is used as the threshold for iris recognition. This method is tested using CASIA1, CASIA2, CASIA3-Interval and MMU1 databases. Experiment results show that this method can extract the binary feature effectively, with faster speed and higher correct recognition rate.
出处
《吉林大学学报（工学版）》
EI
CAS
CSCD
北大核心
2013年第1期198-205,共8页
Journal of Jilin University:Engineering and Technology Edition
基金
中国科学院知识创新计划项目(KGCX2-YW-911-2)
关键词
计算机应用
经验模态分解
改进标准差
虹膜识别
computer application
empirical mode decomposition
improved standard deviation
iris recognition
