Research And Realize Of Segmentation Algorithm Of Fingerprint Image

In the field of biological recognition system,now,automatic fingerprint identification system(AFIS) has been widely accepted and applied although the application of fingerprintidentification system has been very mature and people also have a good understanding of eachstep of the recognition system, but with the development of science and technology and thesocial problems, the study of fingerprint recognition need deeper research to solve the emergingproblems in the application of fingerprint identification. AFIS can be divided into four steps:theacquisition of fingerprint image;preprocessing of the fingerprint image;extract feature,fingerprint matching. In order to reduce the redundancy of unnecessary computation, fingerprintsegmentation needs in front of fingerprint image preprocessing.This article will focus on thefingerprint image segmentation techniques and algorithms to do more in-depth study and make asimulation experiment on each segmentation algorithm.Fingerprint image segmentation is an important part of the preprocessing of fingerprint andthe automatic fingerprint recognition system,It can greatly reduce subsequent processingcomputation and subsequent processing time to separated precisely fingerprint foreground areafrom the background region and will significantly increase the processing speed and quality ofautomatic fingerprint identification system.The main research contents of this paper are asfollows:1. First,This paper presents a kind of segmentation algorithm of fingerprint image based onCMVF features.because The classical algorithm of single feature or simple fusion are difficult tosegment figerprint image effectly for low quality fingerprint image,so the segmentationalgorithm of fingerprint image based on CMVF features is proposed.This algorithm uses threeclassic features,there are the fingerprint direction consistency, average gray, variance. Except,frequency characteristic of fingerprint is used.The four features of the algorithm respectivelyrepresent the gray feature of the image (gray mean and variance), the direction of fingerprint(coherence), and this combining gray, party image,texture represents a new fingerprint textureinformation of the ridge frequency information.The calculation of fingerprint frequency calculatethe distance between crest and adjacent.Because the level of contrast and the noise have no effecton frequency,the segmentation algorithm has better result of segmentation of fingerprint imageswith low contrast or a large number of noise,and skillfully avoide defects for low quality offingerprint image.It is found that the final segmentation results have been greatly improved byadding this feature.2. In addition, this paper also proposes a fingerprint image segmentation algorithm based onAdaboost.Adaboost used in the application of face recognition system and image retrieval.The essence of fingerprint segmentation is the problem of binary classification,and the Adaboostclassifier has a better effect for classification.The method basing on the stage treatment isdivided into three times when segmenting. Firstly,calculating the threshold for block gray variance by the maximum between class variance (Otsu) and segment figerprint image by thethreshold.Otsu takes care of the different gray of different images,so the use of Otsu to calculatethe threshold is better than the experience value.The second step segmentation use the Adaboostclassifier on the results of first segmentation for further processing.In order to improve accuracy,finally,we use morphological for further processing.In this paper,the method of trainingAdaboost classifier can be applied to sensors collecting different fingerprint images and theaverage error rate is low.3. Finally, The simulation experiments are carried out in the Matlab R2008environment,and this paper compares the algorithm of fingerprint image segmentation based on CMVFcharacteristics to the result of Adaboost classifier with CMV features and Adaboost classifiercombining CMVF features.The data shows the algorithm of using the CMVF more precise thanonly using CMV features. So the more detailed description of the fingerprint imagecharacteristics is more contribute to fingerprint image segmentation.But when we addfeatures,we must prevent the redundant features and avoid increasing the time complexity of thealgorithm.