| Fingerprint is a kind of flower patterns formed by mastoid ridge lines that grow on the cutis of finger tail end, which have the characteristics that a piece finger of each one is different and settling invariability for life. Fingerprint has great value for person identification. In the near century, after investigating and exploring unceasingly, people have clear cognition on the characteristic system of fingerprint, classify the fingerprint characteristics and put forward the scientific fingerprint identification thereunder and procedure.At present, fingerprint identification technology (FIT) takes up important station in modern biology identification technologies. Looking from the practicability and the feasibility, FIT can automatically accomplish a series of works, such as fingerprint classification, character distilling, images memory, search, contrast and matching efficiently, swiftly and conveniently. FIT is of the advantages of convenience, high efficiency, impersonality and security, which is more excellent than identification technologies and has been considered as a kind of perfect identification technology.From the 1960's, computer technology enters into fingerprint identification sight. The countries which have developed computer technology, such as the UK, USA, France and Japan, develop various FIT successively and opened up new ways for fingerprint identification. At present, FIT has been applied widely in financial security, digital encryption; electronic business and judicial practice, which will play more and more important effect in future.In recent years, with the rapid development of the digital image-processing and the technology of hardware, FIT has attained largish development. But it also can not meet the need of society development. In the 21st, replacing other identification technologies (such as seal, key, code and signature) by FIT far-rangingly will become a significant topic. The research of FIT has become a focus among the field of model recognition, manipulating an image and computer vision. Usually, FIT uses general characters of fingerprints to carry through species identification. Basing on it, FIT will systemically compare the detail Characters of fingerprints, and then judge that the fingerprints is whether identical or not. FIT include the modules following: Fingerprint image collection, image pretreatment, character pick-up and character matching. Image pretreatment include the steps following: image quality evaluation, image segmentation, image enhancement, fragmentation, binarization and so on.In the identification technology, image enhancement technology is a all-important one. If fingerprint image is not enhanced effectively, fingerprint's features can not be extracted. Many scholars have discussed fingerprint image enhancement modes. Coetzee etc use Marr-Hildreth edge operator to gain the ridge margin image of fingerprint grey chat and bring forward a method that adopts convolution template to enhance the image. Randolph etc bring forward a method that adopts a set of directional filters to enhance the binary image. Sherlock etc bring forward a method that adopts Fourier filter to enhance the image. Hang etc bring forward a method that adopts Gabor filter to enhance the image. Proof by facts, used in finance security, digital encrypts, electronic business fields, these methods can gain preferable effect.However, these methods estimate fingerprint directed graph by analyzing fingerprint image locally, and then enhance the image by filtering. When the source fingerprint image quality is inferior whose resolving power reduces greatly, these methods can't gain local directed graph well and truly and put up effective filtering enhancement. Virtually, fingerprint images which are polluted badly just have low resolution in judicature field. Therefore, the most algorithms at present are difficult to exert effect.At present, in the aspect of dealing with low quality fingerprint, the efficacy of existing FIT is obviously inferior and difficult to satisfy the need of identification. The primary reason is that the most pivotal tache of existing FIT——image enhancement technologies are devised for high quality fingerprint images, which can not gain perfect effect at enhancing lower quality fingerprint images.Aiming at this complexion, in order to advance the efficiency of FIT for dealing with low quality images, this article puts forward a suit of new fingerprint image enhancement algorithms. The algorithms have three portions: The first one is texture information amplification algorithm that is gained by ameliorating the cumulative distribution function of local area histogram equalization algorithm and includes a non-linear fingerprint texture information amplifier, can be used to the faint texture information of the low quality fingerprints, enhance texture contrast; The second one is fingerprint directed graph two steps algorithm. It can be gained by ameliorate classics directed graph algorithm and calculate directed graph with texture amplification image. It calculates valley points' direction after calculating ridge points' direction and has the capability of clearing up noise at a certain extent. Moreover, it adopts histogram arithmetic to calculate valley points' direction and carries through smoothing filtering for directed graph, which can improve the veracity of directed graph ulteriorly. The third one is directed graph correction algorithm. It is gained by ameliorate classics piece directed graph algorithms and can calculate piece directed graph in bigger neighborhood. Compared with the traditional directed graph algorithms, its calculation window is 3 to 5 times bigger than them. When the image is polluted in a large scale, it is more possible to work out exact directed graph by this arithmetic than traditional algorithms and gain more clear-cut, exact fingerprint image by filtering. These algorithms' capacity for repairing directed graph has biggish improvement.A mass of experiments indicate that, it is difficult for existing classical fingerprint image enhancement algorithms to enhance and repair the low quality blocks whose area is bigger than 7×7 pixels, the texture amplification algorithm can repair the blocks whose area is less than 7×7 pixels, the directed graph two steps algorithm can repair the blocks whose area is less than 9×9 pixels, the directed graph correction algorithm can repair the blocks whose area is less than 13×13 pixels perfectly. Compared with traditional algorithms, the image enhancement efficiency of these algorithms is advanced evidently. The algorithms put forward in this article have better effect and reveal obvious advantages when dealing with low quality fingerprint images.Although the algorithms put forward in this article can repair directed graph in a larger local window, they still can not do that macroscopically. Therefore, these algorithms still can not repair and enhance the low quality blocks whose area is bigger than 13×13 pixels perfectly, need to expand the algorithm's treatment window based on fingerprint's characteristics. |
PDF Full Text Request