Cluster based template generation for fast and secure fingerprint matching

Most of the fingerprint recognition systems depend on pair-wise matching of the minutia feature vectors which consist of several spatial properties associated with the minutia points. Due to various non-linear deformations like translation and rotation, minutia based matching is quite a challenging issue. To deal with this problem, most of the techniques in the literature solely depend upon a reference point selection. In this paper, we propose a novel cluster based template generation and matching algorithm which is invariant to translation and rotation of the minutia points. The cardinality of minutia clusters are stored as fingerprint template instead of minutia feature vectors. The stored template reveals no information regarding the spatial pattern of the fingerprint which provides a strong security against information leakage. Finally, a robust match score estimation technique is introduced to increase the matching efficiency. Moreover, the proposed method produces a faster execution. We present the performance on FVC2004 database to show the effectiveness of our method along with a comparative study against some other existing matching algorithms which establishes the efficacy of the proposed method.