Fingerprint Image Preprocessing Algorithm

The fingerprint, as its unique and stable characteristics, has become one of the most widely used features in biometrics. In recent years, the performance has been made a rapid progress on the automatic fingerprint recognition algorithm and the online-fingerprint-capture techniques, including fingerprint capture, images enhance, feature extracting, feature matching and so on.The preprocessing is the most vital step to make the fingerprint feature extracting and matching accurately. The purpose of preprocessing is to decrease the noise in the images, and to improve the clearance of the fingerprint line, and to make a good basis for the next steps. The fingerprint has an obviously texture structured character; meanwhile, the direction field information can reflect its form character directly and abstractly, so the orientation field has a very important research significance. The thesis is focus on the research of the fingerprint images segmentation and the fingerprint orientation field estimate method.According to the basic principles of the fingerprint recognition techniques, the thesis discussed the workflow of the auto-fingerprint- identification system. By collecting and analyzing the research reports and academic papers published at home and abroad in recent years, the two kinds of estimate algorithms of the fingerprint orientation field have been made experiments on and the results have been compared with them. These algorithms include: the neighboring direction template algorithm, the continuously distributed directional image algorithm, the smallest root-mean-square algorithm and the improved algorithm proposed in Literature [31]. In this thesis, a rapidly calculating method of the continuously distributed directional image algorithm and the gradient-based normalized method are proposed. Experiments show that the former gets an obvious effect on the curtailing processing time of continuously distributed directional image algorithm, and the latter is more accurate to estimate orientation fields of the noisy area edge. In this thesis, the experiments of three fingerprint image segmentation algorithms have been done. These algorithms include: the gray variance-based method for fingerprint image segmentation, the direction-based method for fingerprint image segmentation, the segmented method proposed in the literature [43], and so on. According to the experimental results, the different characteristics of several segmentation features have been analyzed. In the end, the major findings are summarized and the direction for further study is provided.