Low Quality Fingerprint Preprocessing Technology Based On Gabor Filter And Two-dimensional Maximum Entropy

With the rapid development of computer and image processing technology,fingerprint extraction and fingerprint recognition technology have been constantlyimproved. The research based on automatic fingerprint identification has been widelyused in people’s work and life, for example, access control systems, identificationsystems and a variety of occasion need to determine the identity. While the fingerprintrecognition technology was used in the people’s life and work, the low-qualityfingerprint recognition technology has also draw attention from the academiaextensively. For example, for the scene of crime, the finger mimeograph, the brokenfingerprint picture and the general performance of the fingerprint extraction, thetraditional recognition algorithm is not effective and with the low correct recognitionrate. In this paper, the study of the low-quality fingerprint image enhancement and ridgesegmentation is followed by:(1)Firstly, the basic concept of the fingerprint, the traditional fingerprint processingmethods and the process of fingerprint processing systems were described elaborately.Including the fingerprint image capture, a fingerprint image production, the fingerprintregion detection, the judgment of the image quality, the direction and frequencyestimation of the ridge, the image enhancement, the image segmentation, the imagebinarization and elaboration, the feature point extraction, the removal of the falsefeature points and the feature match and so on. On this paper, the processes of theenhancement, segmentation, refinement and feature extraction for the low-qualityfingerprint are researched.(2)Described two traditional types of fingerprint image enhancement technology:fingerprint enhancement on the time domain and on the frequency domain. Because theshort-time Fourier transform (STFT) can get the ridge’s frequency and directionparameters quickly and accurately in the fingerprint image processing and estimateregional masks to separate the fingerprint foreground region from the background area,it has a certain advantages at the speed and accuracy. In this paper, the short-timeFourier transform (STFT) method is used to get the ridge frequency and ridge directionparameters, and then the method of two-dimensional Gabor filter is applied to enhancethe low-quality fingerprint. Experimental results show that the method not only has agood effect on the noise suppression, also be better at having the noise remain activelyin the fingerprint ridge line direction such that the ridge lines is more clear and theconnection is more natural and smooth.(3)With method of the fusion of Sobel operator and Snake model method for theedge detection, false edges are suppressed, at the same time, the fractures of the edgeare repaired. Thus, the signal-to-noise ratio and the detection accuracy are improved. It can only play a role of smoothing and noise suppression by simply using theSobel-Snake operator, but can not have the separation of the fingerprint image and thebackground. On this paper, the two-dimensional maximum entropy method is proposedwhich could achieve a low-quality fingerprint segmentation by a given threshold vector.The experiment shows that the two-dimensional maximum entropy segmentationmethod based on the the Sobel-Snake operators is not only more flexibly and simply,but also has the divided ridge relative integrity, clear and has the common false edgeand fracture problem from the fingerprint segmentation solved, thus it has a certainvalue to refer.(4) The improved OPTA algorithm for the fingerprint elaboration is applied.Without changing the shape and direction of the fingerprint ridge line, it can find out thedetail features of the fingerprint. On this basis, it has the fingerprint feature extractionand the removal of the false feature points.