ISODATA聚类在指纹图像分割中的应用  

Application of ISODATA in Fingerprint Segmentation

作　　者：刘汉英[1] 周剑勋 谢晓兰[1] 

机构地区：[1]桂林理工大学信息科学与工程学院,广西桂林541004 [2]桂林泓成橡塑科技有限公司,广西桂林541004

出　　处：《桂林理工大学学报》2013年第3期547-554,共8页Journal of Guilin University of Technology

基　　金：广西科学研究与技术开发计划项目(桂科攻11107014-4);广西教育厅科研项目(200708LX226);桂林市科学研究与技术开发计划项目(20110111-1)

摘　　要：针对指纹图像的特点,研究了传统的指纹分割方法和形态学图像处理方法,将迭代自组织分析算法(ISODATA聚类)应用到指纹图像分割中。首先根据指纹图像块灰度均值、块标准偏差、块灰度对比度和块方向一致性,使用聚类方法划出图像的部分背景块;然后将背景块灰度用背景处灰度平均值代替,使用形态学图像处理方法实现指纹图像的分割。对FVC(国际指纹识别竞赛)指纹库中不同质量的指纹图像进行了测试。结果表明,此算法比传统算法更有效。Fingerprint image segmentation is the first step of automatic fingerprint identication. Based on the characteristics of fingerprint, traditional fingerprint segmentation and morphological image processing are stud- ied, Iterative Self-Organizing Data Analysis Technique Algorithm （ISODATA） and morphological processing are used in fingerprint segmentation. Firstly, the features of each block are computed, such as the average im- age intensity, standard deviation, gray-scale consistency and directional consistency. Some of the background blocks are divided by clustering method. Then the gray background blocks are replaced by the average value. Morphological image processing methods are applied in fingerprint segmentation. The algorithm is tested on the different quality fingerprints of the FVC （ Fingerprint Verification Competition） fingerprint databases. The ex- perimental results show that the present algorithm is better than the tradition fingerprint image segmentation algo- rithms.

关 键 词：迭代自组织分析算法 形态学处理 指纹图像分割 自动指纹识别 指纹特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]