An Improved K-Means Based Method for Fingerprint Segmentation with Sensor Interoperability

Abstract

Fingerprint segmentation is an important step in an automatic fingerprint recognition system. Due to applications of various sensors, fingerprint segmentation inevitably suffers from sensor interoperability problem. K-means algorithm is one solution to address the sensor interoperability problem in fingerprint segmentation. However, the traditional k-means based method does not well deal with the border between the foreground and the background. The over-segmentation of foreground area may appear and some important minutiae are lost. To effectively address the issue, we propose an improved k-means based segmentation method with sensor interoperability called ISKI. ISKI performs the secondary determination to the blocks which have similar distances with the two cluster centers after k-means clustering. The proposed method is applied on a number of fingerprint databases which are collected by various sensors. Experimental results show our proposed method significantly improves the accuracy of segmentation.