In fingerprint verification or identification systems, most minutiae-based matching algorithms suffered from the
problems of non-linear distortion and missing or faking minutiae. Local structures such as triangle or k-nearest structure
are widely used to reduce the impact of non-linear distortion, but are suffered from missing and faking minutiae. In our
proposed method, star structure is used to present local structure. A star structure contains various number of minutiae,
thus, it is more robust with missing and faking minutiae.
Our method consists of four steps: 1) Constructing star structures at minutia level; 2) Computing similarity score for each
structure pair, and eliminating impostor matched pairs which have the low scores. As it is generally assumed that there is
only linear distortion in local area, the similarity is defined by rotation and shifting. 3) Voting for remained matched
pairs according to the compatibility between them, and eliminating impostor matched pairs which gain few votes. The
concept of compatibility is first introduced by Yansong Feng , the original definition is only based on triangles. We
define the compatibility for star structures to adjust to our proposed algorithm. 4) Computing the matching score, based
on the number of matched structures and their voting scores. The score also reflects the fact that, it should get higher
score if minutiae match in more intensive areas. Experiments evaluated on FVC 2004 show both effectiveness and
efficiency of our methods.