摘要
近年来,二维主分量分析(2D-PCA)和离散小波变换作为图像分析的两种有效方法,受到人们的广泛关注。结合以上两种方法,提出了一种多频带2D-PCA虹膜识别快速算法。该算法首先对虹膜图像做预处理,然后将预处理后的图像做2维离散小波变换,取小波系数的两个中频子带作为2D-PCA的输入空间;在训练阶段,求得训练样本输入空间的特征空间并由此得到训练样本的特征向量,形成样本特征库;在识别阶段,计算得到未知样本特征向量;同时为了提高特征向量对图像旋转的鲁棒性,在该阶段进行了基于不同起始角度的归一化处理。最后采用Hamming距离,对未知样本的特征向量在特征库中进行多模板匹配,通过K临法则和阈值法得到识别结果。实验结果验证了所提算法的有效性。
Recently, as tow effective methods for image analysis, 2D-PCA and wavelet transform get extensive attentions. A fast iris recognition arithmetic was proposed in this paper based on 2D-PCA and wavelet. Firstly, we pre-dealt with the image; then applied 2D-PCA on the two third level middle-frequence subband of wavelet coefficient, then got the feature vector by combination and symbol quantify ; finally, we applied multi-templet matching between unknown class sample and feature database,in the same time, in order to increase the robustness against rotation of the original image, we applied a anti-rotation method on the unknown class image;then got the recognition result by K-Neighbor and threshold. The experiment result validates the efficiency of the arithmetic proposed.
出处
《计算机科学》
CSCD
北大核心
2009年第10期280-283,共4页
Computer Science
基金
辽宁省自然基金项目(20072156)
辽宁省教育厅科学技术研究项目(20060486)
南京邮电学院图像处理与图像通信江苏省重点实验室开放基金(ZK207008)资助