Localized Dictionaries Based Orientation Field Rebuild For Incomplete Fingerprint

Fingerprint orientation field clearly shows the global information of fingerprints at macroscopic, such as the basic shape,structure and the direction of ridge line, and preserves largely details information simultaneously. Thus, in practical applications, the correct estimation of fingerprint orientation field is usually the key step of fingerprint identification technology. The proportion of incomplete fingerprints in our country fingerprint files is large. The poor state of finger(exist scars, peeling, dirt, et al.), fingerprint acquire equipment defects and improper press way, will result in the obtained fingerprint image characteristics information missing and nonlinear deformation, and bring the huge challenge to fingerprint identification technology. Existing fingerprint orientation field reconstruction algorithms are hard to achieve satisfactory results on the accuracy and timeliness simultaneously, which limits the application range of fingerprint identification. Thus, the improved incomplete finger orientation field reconstruction method based on local dictionary is proposed in this paper, which aims to improve the accuracy and decrease the time costs of reconstructing incomplete fingerprint orientation field at the same time. This method will promote incomplete fingerprint orientation field estimation algorithm of research and development of fingerprint identification in practical application. The main innovation and improvement of this paper summarizes as follows:(1)On spatial distributions estimation, the determining clustering numbers based k-means++ is instead of traditional k-medoid in this paper.So a faster clustering speed and better clustering effect simultaneously isachieved.(2)On fingerprint pose estimation, extract effective area of incomplete fingerprint orientation filed before pose estimation, which can reduce the invalid area to participate in the "vote", leading to faster and more accurate incomplete fingerprint post estimation.(3)On local dictionary constructing, considering multiple neighborhood priori information instead of a single location improves the diversity of orientation field patches in local dictionary. Candidate orientation field patches obtained by local dictionary query can be closer to the original information in incomplete fingerprint, which can reconstruct the fingerprint orientation field more accurately.Finally, in order to verify the performance of the algorithm in this paper, comparison experiments are implemented on several traditional algorithms and the proposed method about accuracy and timeliness.Overall, Experiment results show the proposed method performs better accurate and less time consuming simultaneously.