Pore-based indexing for fingerprints acquired using high-resolution sensors

With the advent of high-resolution fingerprint sensors, there have been efforts to develop high-resolution fingerprint identification systems. In this paper, we present an indexing method for high-resolution fingerprints using pore-based features. In the proposed approach, the dynamic pore filtering method is employed to extract pores from high-resolution fingerprint images. The extracted pores are treated as keypoints, and a pore descriptor is computed for each of the keypoints. The pore descriptor thus obtained is used as a feature vector for indexing. Finally, a cluster-based retrieval scheme is employed for fast and effective retrieval of the candidate list. Performance of the proposed approach has been evaluated on the Hong Kong PolyU high-resolution fingerprint databases, DBI and DBII and in-house databases, IITI-HRFP and IITI-HRF. The proposed indexing approach achieves an improvement of 67%, 49%, 42% and 28% points in pre-selection error rate (for penetration rate of 10%) over the existing method that employs pore features for indexing on DBI, DBII, IITI-HRFP and IITI-HRF, respectively. Most importantly, our approach provides better performance than the state-of-the-art minutiae-based fingerprint indexing algorithm on DBI and IITI-HRFP, which contain partial fingerprints.