基于场结构的指纹图像细节特征提取算法  

An algorithm of minutia extraction for fingerprint image based on the orientation field and frequency field

作　　者：何洋[1] 欧宗瑛[1] 

机构地区：[1]大连理工大学精密与特种加工教育部重点实验室,辽宁大连116024

出　　处：《石油大学学报（自然科学版）》2004年第3期60-64,共5页Journal of the University of Petroleum,China(Edition of Natural Science)

基　　金：国家"8 63"高科技计划项目 ( 863 3 0 6 ZD13 0 3 6)

摘　　要：根据指纹图像局部区域所固有的方向特性和频率特性 ,提出了方向场和频率场的概念 ,实现了指纹图像与背景区域的有效分离 ,并实现了指纹图像的二值化、细化以及细化纹线修复的后处理。提出了一种基于场结构的快速而有效的指纹图像细节提取算法。该算法依据局部与宏观相结合的原则 ,不但利用了指纹图像的局部特性 ,而且结合了局部四邻区域的关联特性 ,对于脊线方向变化较小的局部区域 ,采取单一方向Gabor滤波的方法 ;对于脊线方向变化较大的局部奇异区域 ,采取多方向Gabor滤波合成的综合方法 ,从而保证了提取指纹细节特征的准确性。实践证明 ,与现有的方法相比 ,此方法具有计算量小、细节特征提取准确的特点 ,可应用于实际的指纹自动识别系统中。According to the intrinsic properties of orientation and frequency,a method for segment of fingerprint from background and the concepts of orientation field and frequency field were proposed. The binary, fineness and postprocessing of fingerprint image were fulfilled. A fast and effective minutia extraction algorithm based on the orientation field and the frequency field was designed. This algorithm utilizes the information in a local specified region of fingerprint images and the associated information of adjacent regions in four directions. Based on the combination analysis of specified region and its neighborhood regions,a single Gabor filter with best orientation and other optimal parameters were selected for the regular regions of fingerprint image. A combined multiple Gabor process was implemented for the non-regular regions,such as the regions existing ridge in abruptly changed direction. The experimental results show that this algorithm requires less computation and can provide better minutia extraction result, compared with other existing algorithms. It can be applied to the practical fingerprint identification system.

关 键 词：生物识别技术 指纹图像 计算方法 GABOR滤波器 自动识别系统 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] Q81[自动化与计算机技术—计算机科学与技术]