期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2006
卷号：6
期号：3A
页码：197-202
出版社：International Journal of Computer Science and Network Security
摘要：In this paper a novel distortion-tolerant fingerprint registration method based on centroid structure vector (CSV) and Vector Line Segment (VLS), is proposed. In this method, minutia features of the query fingerprint are clustered into various clusters and a set of VLS is extracted. Several likely transformations between the query and the template images are estimated by using VLS sets. Then, CSV under each likely transformation is constructed by using the centroids of minutia clusters. The fine transformation is finally determined by the number of VLS pairs together with the similarity level of CSV. Experiment results show that this algorithm is an effective and efficient one for aligning fingerprints with a small number of minutia and heavy distortions. Such situations are often encountered in forensic applications.
