Finger Vein Recognition based on Personalized Discriminative Bit Map

Finger vein recognition is a promising biometric recognition due to its some advantages. For a finger vein recognition system, feature extraction is a critical step for the final recognition. In our previous work, we proposed Personalized Best Bit Map(PBBM) which selected the stable bits from LBP. Although PBBM achieve a better performance, it still contains some useless bits for recognition. In this paper, we propose Personalized Discriminative Bit Map(PDBM) which select much more discriminative bits from PBBM. The bits of PDMB are more discriminative and more effective for the final recognition. In addition, compared with PBBM, the number of bits for matching is reduced, so PDBM can also reduce the computation complexity. Experimental results show that PDBM achieves not only better performance, but also consumes less time for matching.