【摘要】 针对指纹识别中的指纹二值化问题,提出了一种基于4-邻域均值模板和直方图搜索的优化采样方法.定义计算速度快的4-邻域均值模板来区分边缘像素点和中间像素点,同时基于灰度直方图,搜索统计量较多的像素点作为采样点.将优化采样后所得像素值以及4-邻域均值模板计算值作为训练集,通过支持向量机完成指纹二值化、FVC2004数据库验证以及与相关算法的结果比较.结果表明,所提出的算法分类精度更高,对边缘像素点处理更精准,同时计算速度更快.更多还原

【Abstract】 Fingerprint,as a biometric characteristic is often used in many fields.Aiming at the binaryzation of fingerprint image,this paper proposes an optimal sampling method based on4-neighborhood mean template and histogram search algorithm.The fast computation template with4-neighborhood is defined to distinguish between edge pixels and intermediate pixels.Meanwhile,the pixels with high values in histogram are selected as the sample points.Then,the training set which includes the pixels and value of the 4-neighborhood mean template is selected by optimized sampling.Finally,SVM based on training set is used to perform the binaryzation of fingerprint image.The experiments are executed on the FVC2004 database,and compared with the results of the relevant algorithms,the new algorithm can achieve higher classification accuracy and faster calculation,as well as the more accurate processing on edge pixels.In this paper,the binaryzation algorithm of support vector machine based on optimized sampling has made a useful attempt for the research of fingerprint image recognition.更多还原