Fingerprint Image Preprocessing And Feature Extraction Algorithm

With the rapid development of high-tech industry in recent years, biometrics has been widely used in many regions, such as financial security, access control system, digital encryption and E-commerce, etc. Many advantages exit in fingerprint identification technology, such as the convenience of image acquisition, ease of hardware realization, maturity and stability of algorithms, So it has earned broad concern and been used extensively.Based on the existed researches, this thesis mainly researches the preprocessing and features extracting algorithms of fingerprints. In the beginning of this thesis, the history and conceptions in each steps of the entire fingerprint identification system is introduced. The main tasks include the followings:1. Pre-processing of fingerprint image. Firstly, the collected fingerprint images have been normalized by the method, which perform excellently, especially on dim image. And then, we raise a new segmentation algorithm based on statistical characteristics of fingerprint and apply it. About the enhancement, we bring three enhancement algorithms - Gabor filter, homomorphic filter and directional filter. Meanwhile, it is applied to enhance fingerprint that is the combination of homomorphic filter and directional filter. At last, the adaptive threshold algorithm has been used to binary, and the morphology to thin before corresponding post-processing respectively.2. Feature extraction of fingerprint image. We extract minutiae by the minutiae detection template, and recognize singulars (center and triangular points) using back-propagation neural network based on grayscale fingerprint image. After all the features being extracted, we use the ridge line tracing algorithm to judge these features, and then remove pseudo-features.We complete these algorithms mentioned with matlab on PC in this paper. The results prove that our algorithms proposed show excellent performances on image processing compared with others.