指纹图像的预处理及其改进算法

【作者】 王家隆；

【导师】 郭成安；

【作者基本信息】 大连理工大学 ， 信号与信息处理， 2003， 硕士

【摘要】 自动指纹识别是本世纪六十年代兴起的、利用计算机取代人工来进行指纹识别的一种方法。近年来，随着计算机技术的飞速发展、低价位指纹采集仪的出现以及高可靠算法的实现，更使得自动指纹识别技术越来越多的进入到人们的生活和工作中，自动指纹识别系统的研究和开发正在成为国内外学术界和商业界的热点。相对于其它生物特征鉴定技术，例如语音识别及虹膜识别，指纹识别具有许多独到的优点，更重要的是它具有很高的实用性和可行性，已经被认为是一种理想的身份认证技术，有着十分广泛的应用前景，是将来生物特征识别技术的主流。 本文对基于细节特征点的指纹图像预处理、特征提取方法进行研究。 指纹图像预处理是指纹自动识别过程的第一个处理环节，它的好坏直接影响着自动识别系统的效果。通常预处理有滤波、方向图的求取、二值化、细化等几个步骤。 本文采用最小均方估计算法计算指纹图像的块方向；利用基于方向图的动态阈值指纹图像二值化方法可以得到较好二值化图像；细化部分本文提出了改进的并行细化模板，保证细化结果的同时大大减少了细化迭代次数；本文提出一种适合于串行机实现的并行细化算法，该算法可保证在不改变原图像的细化结果的条件下，大幅度提高运算速度；在特征提取部分，以指纹最内层弧的顶点作为中心点，统一了中心点的定义，可以比较准确地提取出指纹中心点。 本文在PC机上用Visual C++编程实现了论文中提到的所有算法，实验结果表明，采用这些算法很大程度地提高了指纹图像处理的效果。更多还原

【Abstract】 AFIS(Automated Fingerprint Identification System) is started from 1960’s, it has replaced conventional method to recognize fingerprints with computers. APIS is applied in daily life more and more with the development of computer technology, the appearance of cheap collector and the effective algorithms. APIS is becoming the research center of academic and commerce. It is superior to the other biological feature with its special virtues; the most important is its practicability and feasibility. It has been considered the main method of the biological feature recognition in the future.The preprocessing and feature recognition of APIS based on minutia features are discussed in this paper. After the research and the comparisons on the exist algorithms, some new algorithms are proposed.The preprocessing is the first step of APIS, it is the most important step. It includes Filter, Orientation, Binarization, Thinning and so on.The least mean square is used to get the blocks orientation; get the binary image with the binarization method of fingerprint images based on orientation and dynamic threshold; in the thinning, an improved thinning algorithm is proposed here in order to overcome the shortages of the existing thinning algorithms, the new templates can significantly reduce the scanning iterations and speed up the thinning process while keeping the same thinning results; the image parallel thinning algorithm for serial computer is suitable for all kinds of parallel template thinning algorithms, it improve the computation efficiency of image parallel thinning on serial computers based on a coding scheme for each 4x4 binary pixels; in the features extraction, the center of fingerprint is defined as the point where the curvature on ridge is maximum, so the definition of the center of fingerprint is unified.All algorithms are implemented on PC by VC++. The results are given in the paper demonstrate that the new methods can enhance processing effects of fingerprint images.更多还原