Fingerprint identification as a biometric is becoming one of the most reliable personal authentication technologies due to the uniqueness and immutability of fingerprints.Now most of the automatic fingerprint identification systems are realized on the PC-based platform. But the newly developed market for civil application has demanded more for system's portability and convenience. So the trend of miniaturizing and embedding the system for automatic fingerprint identification become more and more distinct.The algorithm of automatic fingerprint identification system can be decomposed into the following three fundamental parts: (i) fingerprint image preprocessing, (ii) fingerprint feature extraction, (iii) pattern matching. The fingerprint image preprocessing is the first phase. Only after the preprocessing is performed perfectly, could the subsequent phases be reliable and satisfying.In this thesis, the theory of fingerprint identification technology and the algorithm about image processing are studied and a set of preprocessing algorithms isdesigned, which fit the fingerprint image and may be realized in DSP platform. The research lays the foundation for the subsequent work. Through comparison and experiment, a set of preprocessing algorithms based on ridge direction was proposed, which include calculation and correction of point directional map and block directional map, oriented filtering, directional binarization, thinning. Based on the features and performance of DSP processor, some improvement has been made to optimize further the algorithms.Experimental results show that the optimized preprocessing algorithm can improve the fingerprint image quality efficiently, extract the useful information exactly, and be easy to be realized on DSP. |