Reference point detection for arch-type fingerprint using multi-scale Gaussian filter in orientation field

초록

Reliable reference point detection is still a challenging problem for arch-type fingerprints. Proposed in this paper is a robust reference point detection method for arch-type fingerprints based on pixel-wise orientation field using multi-scale Gaussian filter. Firstly, a pixel-level orientation field is computed using multi-scale Gaussian filter. Secondly, a symmetry line and a angular gradient field are derived from the high-resolution orientation field, then, the local maximum points long the symmetry line are considered as candidate reference points. The final reference point is determined by utilizing the symmetrical and topological structure. A set of arch-type fingerprint images is selected from FVC 2000 db1a for evaluating the proposed method and robust and accurate experimental results are obtained.