An application of linear SVM to fingerprint image segmentation

摘要

An exact fingerprint image segmentation improves the accuracy of automatic fingerprint identification system and reduces the processing time of follow-up process. A novel fingerprint image segmentation method based on linear Support Vector Machine (SVM) is proposed to overcome the shortcomings of present approaches shown in previous literatures. Feature vectors are constructed by Fourier spectrum energy ratio and gray contrast extracted from sub-blocks of the fingerprint image. Then these feature vectors are classified by linear SVM, morphological operations are performed to realize fingerprint image segmentation in the end. Experimental results show that this method is precise and reliable compared with other approaches.