一种基于子块图像互相关的虹膜识别方法

虹膜识别是一种新兴的生物特征身份识别方法,在场所或资源的安全控制等方面具有重要的应用价值。该文提出了一种基于子块图像互相关的虹膜识别方法。在特征提取时,把图像划分为许多子块,计算两幅虹膜图像的对应子块间的归一化相关系数,并把最小的几个归一化相关系数去除掉,然后求得剩下的归一化相关系数的均值及方差,作为两幅虹膜图像的联合特征。这能够有效地减小图像强度的局部畸变对虹膜识别的影响。为了实现模式的非线性划分,判别函数采用径向神经网络来构造。最后,进行了小样本实验,实验结果表明,这种基于子块图像互相关的虹膜识别方法是可行的。

An Iris Recognition Method Based on Sub-image Correlation

Iris recognition is an emerging biometric technology.It is of great value in the security for certain places or resources.An iris recognition approach based on sub -image correlation is presented in this paper.For the feature extraction,the whole image is divided into many smaller blocks and then the normalized correlation coefficients between the every two corresponding sub-images are computed.The mean value and standard variance of them are thus obtained to be the joint characteristics of two iris images after some smallest normalized correlation coefficients are discarded.The approach helps to decrease efficiently the influence of local change of image intensity on iris recognition.The judgement function is constructed by radial neural net in order to provide nonlinear discrimination.At last,the experiment with small samples is made.Its results show that the iris recognition method based on sub-image correlation is feasible.