Research On The Fingerprint Image Segmentation Algorithm

Today in the 21st century, along with the development of information technology and computer network technology, anti-risk capability of the traditional personal identity recognition methods based on identifier or given knowledge cannot satisfy the application requirement. So, it's one of key issues in the authentication field that how to improve the reliability of identity recognition.Because of the high-security, non-substitute and high-distinguishability, biometrics identification technique becomes key technology to identify personal identity in the modern information words. Automatic fingerprint identification technique is regarded as one representative biometrics identification technique by domestic and overseas researchers. In many years, researchers have investigated technical issues in automatic fingerprint identification technique in detail, brought forward the basic technology system of automatic fingerprint identification technique and achieved plentiful research harvest. However, many key issues have not been settled entirely in automatic fingerprint identification technique, limit application field and scope of automatic fingerprint identification system.In these key issues, how to process the low-quality fingerprint images is one of key issues, how to segment low-quality fingerprint images determines subsequent fingerprint image process performance and has become one difficult issue we should figure out urgently. Consequently, the paper chooses the issue as primary research domain and brings forward two kinds of fingerprint segmentation algorithm faced with low-quality fingerprint images.After researching on several kinds of fingerprint image segmentation algorithm, the paper introduces the multi-agent system into the fingerprint image segmentation and proposes one kind of self-adaptive fingerprint image segmentation algorithm based on multi-agent system. The algorithm segments the fingerprint images utilizing the competition and dilatation between agents. When the MAS is counterpoised, the algorithm have segmented the fingerprint images accurately. Now, the distributing results of detectors are the fingerprint image segmentation results. Contrastive experiment results show that the algorithm can segment the fingerprint image exactly and self-adaptively. The algorithm proposed in the paper has better anti-noise ability and robustness to the noise in the fingerprint image. In the processing process to many kinds of low-quality fingerprint images with low contrast, high background noise, the segmentation results are more accurate, reliable and effective.The paper introduces the voting decision system into the fingerprint image segmentation processing and proposes the fingerprint image segmentation algorithm based on multi-factors voting decision system. Firstly, the algorithm selects block image gray variance, gray mean, block orientation variance mean and point orientation concentration as the segmentation factors. Secondly, the algorithm segments the fingerprint images using the single-factor, double-factors and multi-factors respectively and obtains the segmented fingerprint image. Lastly, the algorithm acquires the final segmented fingerprint image utilizing the voting decision system. Contrastive experiment results show that the algorithm proposed in the paper has better anti-noise ability than these algorithms based on one segmentation factor. The segmentation result is more perfect to the low-quality fingerprint images because the phenomenon of inaccurate segmentation or excessive segmentation is decreasing. The algorithm proposed in the paper has better adaptivity and robustness for the fingerprint image quality than other algorithms.