Research On Low-Quality Fingerprint Enhancement Algorithm

Automatic fingerprint recognition is a more sophisticated biometric recognition technology, which is extensively applied in the field of entrance control, human-computer interaction and so on. But with the development of information society, the present methods have not met the urgent need.Therefore, in recent years, a large number of research agencies are actively involved in research of the key technologies in this field in order to improve the performance of automatic fingerprint recognition system. Fingerprint enhancement is particularly important in the fingerprint identification process, and if this part was't dealt with, it would be difficult to obtain good results by improving following minutiae extraction. Good fingerprint enhancement alogrithms are the basis of improving the performance of system. In this thesis, low-quality fingerprint image is the object of the study, and an in-depth research is done on low-quality fingerprint image enhancement. The main tasks are as follows:(1) The current commonly used three kinds of fingerprint enhancement algorithms, based on the root filter, traditional Gabor and STFT, are analyzed and simulated. By comparing the performance of different algorithms, it is concluded that fingerprint enhancement algorithm based on STFT has the highest accuracy, therefore, it is selected as the reference object.(2)A low-quality fingerprint image enhancement algorithm based on STFT and Gabor is presented. The algorithm simultaneously obtains the orientation field and frequency field through STFT, and then use Gabor filter, the optimal joint resolution in both spatial and frequency domain, to fulfill the filtering task. Simulation results show that the presented algorithm has a better enhancement result and faster execution speed with respect to the reference algorithm.(3)A directional Gabor filter bank is constructed, and the subfilter in each direction is convolved with the fingerprint image. According to the orientation field through STFT, integration of sub-image is finished, resulting in the enhanced image. Experimental results show that the algorithm is simple and has an better enhancement result.(4)NSCT is introduced in the process of low-quality fingerprint enhancement, the transform is suitable for denoising and enhancement of the texture image. At the pre-processing stage of fingerprint recognition, nonsubsampled Contourlet transform is added, and then conventional fingerprint enhancement algorithms are used. The experimental results show the quality of enhancement can be further improved.At last a demo system of automatic fingerprint recognition is implemented.