Segmentation of fingerprint images — A composite method

摘要

In this paper the limitations of the segmentation of FP images using the directional criterion as suggested in Ref. (1) are pointed out. The equation used for the direction computation becomes undefined when the input image has perfectly uniform regions. To overcome this difficulty a composite method of segmentation is suggested which uses the variance criterion wherever the directional method fails. The failure of the directional method is detected by the low gray variance property of the uniform regions. The composite method takes advantage of both the directional and variance methods. This composite method is tested for a large number of FP images from different sources, and is found to be effective and robust. The results of segmentation of FP images using the composite method and the two methods separately are presented. The desirable features of a segmentation method are also given.