指纹图像分析及特征提取
文献类型：学位论文
| 作者 | 师忠超 |
| 学位类别 | 工学博士 |
| 答辩日期 | 2005-05-01 |
| 授予单位 | 中国科学院研究生院 |
| 授予地点 | 中国科学院自动化研究所 |
| 导师 | 王阳生 |
| 关键词 | 指纹分割 Fingerprint Segmentation Coherence-enhancing Filter Chebyshev Approximation 一致性增强滤波 契比雪夫逼近 特征提取 算法融合 |
| 其他题名 | Fingerprint Image Analysis and Feature Extraction |
| 学位专业 | 模式识别与智能系统 |
| 中文摘要 | 随着人们对社会安全要求的增加，基于生物特征识别的智能身份鉴别方法逐渐受到广泛的关注。指纹识别技术是到目前为止各种生物认证技术中发展最早、应用也最广泛的技术，而且随着需求的增加，人们对指纹自动识别系统性能的要求也不断提高。虽然经过几十年的理论研究，已经存在着多种识别算法和实际应用系统，很多研究者们也在指纹分类、指纹增强、特征提取以及匹配算法方面发表了大量的文章，但是仍然存在着许多值得研究的问题。本文的工作是在我们原型系统的基础上，对系统存在的一些问题进行了分析，并对指纹图像分割、增强以及特征提取等问题进行了一些研究，取得了一些研究成果。论文的主要工作和贡献如下： 1) 针对指纹图像的特点提出了一种新的低质量指纹图像分割方法。在该方法中，我们对原始指纹图像进行分析评价，首先通过两种统计特征定位出噪声信号的灰度分布区域，然后对原图像进行非线性变换，根据变换后的统计信息结合灰度和纹理的信息对图像进行分割。实验结果表明，该方法是一种有效的分割方法，尤其是在处理受污迹和汗渍影响的低质量图像时，效果更加明显； 2) 对指纹增强算法进行了研究，首先提出了一种改进的 Gabor 滤波算法，实验结果表明该算法要好于 Jain 的 Gabor 算法。但 Gabor 算法容易造成细节点的类型变化以及位置的移动，而且在图像较差区域由于其纹线宽度以及方向图精度的影响，效果不甚理想。因此，我们对基于一致性增强扩散的指纹增强方法进行了一些深入的研究，提出了一种新的方法来提高算法的效率，并且针对增强图像的特点，提出了一种基于图像地形特性的指纹二值化方法。实验结果表明，这种算法可以有效地增强图像，它不但可以有效地连接中断的纹路，而且可以正确地分离对比度较小的纹线，并且在奇异区域的效果要远好于 Gabor 算法的效果； 3) 提出了一种结合图像质量信息和结构信息来去除伪细节点的特征后处理算法。在该算法中，我们首先利用细化图中细节点之间的结构信息去除可能的伪细节点，然后利用图像质量信息来进一步处理保留下来的细节点，以得到最后的细节点。实验结果表明，该算法可以有效地消除伪节点，以得到最后的细节点。实验结果表明，该算法可以有效地消除伪细节点，尤其在处理质量较差图像时效果更加明显； |
| 英文摘要 | With the increasing requirement for security, biometrics based personal identification methods have received extensive attention. Fingerprint recognition technique is the earliest one and is applied most widely now. With the increasing requirement, people also want the automated fingerprint recognition system become more reliable. After the academic research for decades, there are some recognition algorithms and practical application systems and plentiful papers published, but there are still some problems remained to be valuable for research. In this thesis, we analyze some problems in our prototype system and investigate some of key issues in fingerprint image processing and feature extraction, including fingerprint image segmentation, enhancement and feature extraction. The main contributions of our work reported in this thesis are as follows: 1. A new fingerprint image segmentation algorithm for low quality fingerprint is proposed according to the characteristic of low quality fingerprint. In this method, the original fingerprint image is evaluated and the gray distribution region of noise signal is found out by two statistical features firstly, and then original image is transformed nonlinearly, finally, the image is segmented by combining new gray and texture information. Experimental results have shown that this method is effective and works better than others in processing the low quality image which is polluted by smear or sweat; 2. We make some research in fingerprint enhancement algorithms. Firstly, anew Gabor-based algorithm is proposed. The experimental result haveshown that its performance is better than that of Jain’s. But Gabor-basedalgorithm may result in the change of the type and displacement ofminutiae, and it cannot work well in the low quality fingerprint area due tothe low precision of line width and orientation map. So we make a deeperstudy on the fingerprint enhancement algorithm which is based on thecoherence enhancing-based diffusion, and a new method is proposed toimprove the algorithm’s efficiency. A topographical characteristic-basedbinarization method is also proposed according to the characteristic ofenhancement image. Experimental results have shown that this algorithmcan enhance image effectively, it can not only connect broken line, but alsoseparate the lines with low contrast correctly, and it has better effect insingular area than Gabor-based algorithm; |
| 语种 | 中文 |
| 其他标识符 | 200218014603217 |
| 源URL | [http://ir.ia.ac.cn/handle/173211/5855]  |
| 专题 | 毕业生_博士学位论文 |
| 推荐引用方式 GB/T 7714 | 师忠超. 指纹图像分析及特征提取[D]. 中国科学院自动化研究所. 中国科学院研究生院. 2005. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。