As a well-developed branch of biometrics recognition, fingerprint recognition has a more and more extensive application field because of its great importance. Social activities have such an increasingly demand on the identification and verification that they put forward a more and more restrict requirement on the performance of fingerprint recognition system. Thus the quality of the input fingerprint image becomes a critical factor. There are by far few research on low quality fingerprint image preprocessing, which is the main object in this paper. We focus on several aspects such as fingerprint segmentation, gray level restoration, image enhancement and binarization, and receive some valuable result. The main contribution of this thesis include following issues: (1) We propose a new segmentation method based on region of interest. This method inducts gray level evaluating into the blockwise statistics and apply segmentation to the statistical features. It is proved to have good effect on reducing the influence by sweat spot, false trace and other noise. (2) We improve the flow of enhancement and the algorithm implement. We combine normalization and histogram equalization to restore the gray level information and the weak texture. And we also adopt the Gabor filter in enhancement with the parameter adaptively selected. (3) In order to deal with the low quality enhancement result, we propose a method of binarization with the fingerprint texture approximated by Chebyshev polynomial and deciding whether ridge or valley by means of topographic method. It give the good result especially on the weak texture. (4) We propose a more flexible and reliable flow in low quality fingerprint image preprocessing. According the experiment result of the whole database, our research work are proved to be helpful.
修改评论