Applications of Gabor Functions on Fingerprint Representation, Enhancement, Verification, and Identification

• Main Authors: Chih-Jen Lee, 李志仁
• Other Authors: Sheng-De Wang
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/78227525989803394566

博士 === 國立臺灣大學 === 電機工程學研究所 === 89 === This thesis describes how to take advantage of Gabor filters to the applications of fingerprint, such as enhancement, singular point detection, pattern extraction, verification, and identification. Based on the global and local information, we develop two kinds of Gabor filter-based approaches to extract the fingerprint features. To entirely analyze the influence of the parameters of Gabor filters for the global approach, we transform a local fingerprint image into a complete Gabor basis functions (GBFs) and find that the radial frequency is the most important parameter for fingerprint recognition. Moreover, we also find that the principal GBF, which has the maximum response over the complete GBFs, is sufficient to represent the ridge orientation and frequency of a local fingerprint image and the corresponding response can be used as an index of both the clearness of ridge structures and the area of fingerprint pattern. According to the ridge structures of the core region or those of whole fingerprint, we extract a set of principal GBFs to form a template. Then the core point detection and fingerprint matching can be performed directly from gray-scale image based on the largest responses. We also develop a fingerprint enhancement algorithm with adaptive Gabor filters to obtain very clear ridge structures. At last, we evaluate the performances of the global and the local Gabor filter-based features by various experiments of various system designs, such as verification and identification. Through various comparisons, in conclusion, the local approach is superior to the global from all situations, including accuracy, speed, and memory space.