摘要
提出了一种基于不可分小波和线性判别分析(LDA)的掌纹特征提取算法.该算法首先对掌纹灰度图像进行二层小波分解,保留图像的低频段,这样一幅128×128的掌纹图像经过上述小波分解后得到一幅32×32的掌纹子图,图像的维数显著降低,并且减少了光照这一奇异信息变化对识别效果产生的影响,然后利用LDA方法进行特征提取.针对PolyU掌纹库的识别结果表明,该方法在识别时间,识别率等方面都有其独特的优越性.
A new feature extraction algorith was introduced, which based on non-separable wavelet and linear discriminant analysis（LDA）. Firstly, palmprint images were representation by the lowest resolution subbands after 2-level non-separable wavelet decomposition,in this way, a palmprint image with a resolution of 128 × 128 after the wavelet decomposition got a subimage of 32 × 32 resolution, then the image dimension significantly reduced, what＇s more, it reduced the effects of illumination of this singular information on recognition results. Secondy, the linear discriminatnt analysis was used for feature extraction. The experimental results based on PolyU palmprint database showed that our method was superior to other methods in terms of recognition accruacy and efficienty.
出处
《湖北大学学报（自然科学版）》
CAS
北大核心
2008年第3期249-252,共4页
Journal of Hubei University：Natural Science
基金
湖北省自然科学基金(国际合作)(2007CA011)资助项目
关键词
不可分小波
线性判别分析
特征提取
non-separable wavelet
linear discriminant analysis
feature extraction
