基于小波变换的掌纹特征提取  

Palmprint feature extraction based on wavelet transform

作　　者：苏晓生[1] 林喜荣[1] 丁天怀[1] 周云龙[1] 宋炯[1] 

机构地区：[1]清华大学精密仪器与机械学系精密测试技术及仪器国家重点实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2003年第8期1049-1051,1055,共4页Journal of Tsinghua University(Science and Technology)

摘　　要：掌纹中不同区域的纹线方向和空间频率代表着掌纹图像内在的特征。为了提取该特征,提出了一种基于原始灰度图像的小波变换的新算法。由于该算法直接对原始的灰度图像进行处理,而无需对图像进行预处理(例如图像增强、滤波、区域分割、二值化、纹线细化等),因此它与传统的细节特征识别方法相比大大减小了计算消耗。对一个小的掌纹图像数据库,使用K近邻(K-NN)的分类器对算法进行了实验,得到了很高的分类正确率,验证了算法的有效性。Ridge orientations and ridge spatial frequencies in various regions of the palm represent the intrinsic characteristics of a palmprint image. Palmprint features were accurately extracted using an algorithm based on wavelet transforms of the original gray scale image. The features were extracted directly from the gray scale palmprint image without preprocessing (i.e. image enhancement, filtering, region segmentation, binarization, ridge thinning, etc.), and hence the proposed algorithm requires less computational effort than conventional algorithms based on minutiae features. The algorithm can achieve high recognition rates when a test is on a small palmprint database using the Knearest neighbor (KNN) classifier.

关 键 词：图像处理 生物特征识别 小波变换 掌纹识别 特征提取 灰度图像 计算机图形学 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] D918.91[自动化与计算机技术—计算机科学与技术]