Skip to Main Content
This paper describes a new fingerprint indexing approach based on vector and scalar features, obtained from ridge-line orientations and frequencies. A carefully designed set of features and ad-hoc score measures allow the proposed indexing algorithm to be extremely effective and efficient, as confirmed by the results of extensive experiments. The new method markedly outperforms competing state-of-the-art techniques over six publicly available data sets. Furthermore, it can scale to large databases without losing accuracy: on a standard PC, a search over one million fingerprints takes less than 1 s.