【作者】 郭玉兵；

【导师】 胡咏梅；

【作者基本信息】 山东大学 ， 生物医学工程， 2011， 硕士

【摘要】 近年来,随着计算机技术和人工智能技术的发展,生物特征以其唯一性、稳定性、较高防伪性和随身携带性等优点,被越来越多的应用于身份识别领域。基于指纹的身份识别方式与其他生物特征识别方式相比,不仅具有安全方面的优势,还具有很高的稳定性、实用性和可行性,因此被广泛应用于社会安全、金融安全、公司考勤、门禁管理、机场海关等领域。一套完整的指纹识别算法一般包括图像预处理、特征提取以及特征匹配三个步骤。图像预处理可以削弱指纹图像中的噪声,提高指纹纹线的清晰度,是特征提取和特征匹配的基础,其结果的好坏能够严重影响整个指纹识别系统的准确率。目前,研究学者已经提出了很多针对指纹图像预处理的算法,但是这些算法对低质量指纹图像的预处理效果仍不够理想,所以对指纹图像预处理算法的研究依然是生物特征识别领域的研究热点。本文在查阅和吸纳已有的指纹识别研究成果的基础之上,重点研究了指纹图像的预处理算法,主要包括指纹图像的分割、增强、二值化和细化算法,做了以下几个方面的工作：(1)在指纹分割方面,为了给后面的各种处理提供统一规格的图像,首先介绍了图像归一化算法,对指纹图像的灰度进行变换。然后比较了基于方差和方差梯度的分割算法以及基于方向一致性的分割算法的优劣,并提出了一种基于梯度向量模的分割算法,实验证明该算法可以取得良好的分割效果。(2)在指纹增强方面,首先介绍了最常用的指纹方向图和频率场的计算方法,然后分析了目前应用广泛的Gabor滤波增强算法的局限性,介绍了基于分解Gabor滤波器的增强算法,并对分解Gabor滤波增强算法做了改进以进一步缩短增强处理的时间。(3)在指纹二值化方面,介绍了固定阈值二值化法和基于方向信息的二值化方法,然后根据两种方法的优缺点提出了一种将两者结合的二值化方法,并介绍了一种二值化后处理方法。(4)在指纹细化方面,分析了最常见的快速细化算法和改进的OPTA细化算法的不足,然后介绍了基于8邻域的查表细化算法,最后通过实验证明该算法能够取得更理想的细化效果。最后,对全文进行了总结,指出需要改进的地方,并提出了进一步研究的方向。更多还原

【Abstract】 In recent years, biological features have been widely used in the field of personal identification with the development of computer technology and artificial intelligent technology, in that they have the advantages of uniqueness, stability, higher security and portability. Compared to other biometric identification methods fingerprint identification not only has the security advantages, but also has higher stability, practicality and feasibility. So it is widely used in social security, financial security, company attendance, entrance control, airport, customs and other fields.A complete set of fingerprint recognition algorithm generally includes three steps: image preprocessing, feature extraction and feature matching. As the basis of feature extraction and matching, the preprocessing on fingerprint images can reduce the noise and improve the clearance of fingerprint ridge, so the result of preprocessing can severely affect the accuracy of identification system. Currently, researchers have proposed many methods for fingerprint image preprocessing, but the processing results of images with low quality are not ideal. Therefore, the fingerprint image preprocessing algorithm is still a hot research field of biometric identification.By summarizing and digesting the existing researching, this paper focuses on the fingerprint image preprocessing algorithms including image segmentation, enhancement, binarization and thinning. This article mainly completes the following works:(1) In the step of fingerprint segmentation:firstly, we introduce the image normalization algorithm to transform the gray scale, in order to provide unified specifications for behind steps. Then we compare the advantages and disadvantages of the segmentation algorithm based on variance and its gradient and the algorithm based on orientation coherence. Finally, we propose a new segmentation algorithm and experimented to prove the algorithm can achieve ideal segmentation results.(2) In the step of fingerprint enhancement:firstly, we describe the most common methods for getting orientation and frequency field of fingerprint image, and then discuss the Gabor filter based fingerprint enhancement algorithm and introduced the separable Gabor filter based algorithm. Finally, in order to shorten the processing time, we propose an improved method on the basis of the separable Gabor filter based algorithm,(3) In the step of fingerprint binarization:We discuss the binarization algorithm based on fixed threshold and the one based on orientation information. Then a combined binarization method is proposed according to the advantages and disadvantages of the two above methods. At last, a post-processing method is introduced.(4) In the step of fingerprint thinning:Firstly, we discuss the disadvantages of quick thinning algorithm and improved OPTA thinning algorithm. Then we introduce an index thinning algorithm based on eight neighborhood points and experimented to prove the algorithm can achieve ideal thinning results.Finally, the major findings of the study are summarized and limitations are pointed out to give the directions of future work.更多还原