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 Knearest neighbor (KNN) classifier.
出处
《清华大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2003年第8期1049-1051,1055,共4页
Journal of Tsinghua University(Science and Technology)
