Fingerprint Image Segmentation Based on Support Vector Machine

Exact segmentation of fingerprint image is very important for fingerprint singular points and minutiae features extraction. In this paper, a method for fingerprint image segmentation is proposed based on Support Vector Machine (SVM). The fingerprint image is broken into 16*16 prospects blocks and background blocks. The block average gray, block gray variance, block contrast and the largest peak of non-DC amplitude are used as the feature inputs. An optimization filtering method is used to obtain the kernel function parameters and the error penalty factor. With less training samples, a classifier with good generalization performance is gained. After post-processing using morphological method, the algorithm is simulated with the FVC2002DB-4B fingerprint database. The result of the simulation shows that the correct rate increase to 99%. Both the theoretical analysis and the experimental results indicate the validity of the proposed method.