Evaluation Of Low Quality Fingerprint Images And Detection And Repairing Of Crease

Biometrics is one of recognition techniques to personal identification by using human physiological characteristics. Personal biological characteristics has the virtues with unique and irreproducibility. Today, with the rapid development of computer and information technology, biometric technology, which can meet the personal identification and security requirements, has become a widespread and effective authentication technology. Among them, automatic fingerprint identification technology has become the most mature and most widely used technology in biometric technology. Automatic fingerprint identification technology has high accuracy, high reliability, low cost and many advantages. Because there are usually present a lot of low quality fingerprint images in database, the reliability of automatic fingerprint identification system are seriously reduced. In order to improve the performance of automatic fingerprint identification system, low quality fingerprint image quality assessment and crease detection and repairing methods are studied in this paper.(1) By combining the spatial and the frequency domain quality indices, comprehensive assessment of the fingerprint quality can be achieved. Spatial quality evaluation indices reflect the local ridge structure information of fingerprint. Spatial assessment includes the foreground fingerprint, deflect, wet and dry, ridge continuity and clarity evaluation. Frequency domain index reflects the ridge and valley structure information of fingerprint image, and measures the energy concentration in the frequency domain. According to the evaluation of each quality index, we can get the possible existence of quality problems and give the suggestion to improve the low quality fingerprint. This paper selects five quality indices, as well as using a nonlinear fusion weighting to achieve comprehensive assessment of overall quality. Experimental results on FVC2004 DB1_A, DB2_A and DB4_A show that fingerprint identification system stability is able to substantially increase with 10% low quality images has been excluded.(2) Based on the minutiae information, this paper proposes some approaches to low-quality fingerprint images creases detection and repairing. The method has the advantages with high accuracy and real-time processing, and there are relatively less results in new pseudo minutiae or false ridges. Based on minutiae information of the fingerprint, we not only can detect the position of crease, and breakthrough the limitations of crease direction and shape, but also can detect and repair the pseudo-ridge and fracture in thinning image. According to the length of the fracture and the direction of deflection in thinning image, crease can be divided into three categories. Region ridge growth method is proposed to repair the small length and deflection fracture ridge. For long distance and big deflection fracture ridge, combining of the ridge reference method and the ridge tracing method are introduced to accurately repair crease. Here, ridge tracing method also is used for detecting and repairing other two common pseudo ridge structures. Experimental results show that the algorithm can accurately detect and effectively repair fracture ridges in thinning image with crease, it presents that the method can effectively improve the performance of automatic fingerprint identification system.