Application Of Ordinal Feature For Fingerprint Identification And Cancelable Templates

Fingerprint feature representation is a key actor for fingerprint identification. Commonly used fingerprint features are fingerprint minutiae and imaged-based feature. Although fingerprint minutiae are considered to be the most discriminative features, the original fingerprint image can be restored from the fingerprint minutiae template, which would threaten the privacy of users. And three advantages can be listed to support the image-based approaches: 1) The first is the better capability of image-based approach in dealing with low quality images. 2) The fusion between minutiae-based approach and image-based approach outperforms the minutiae-based one. 3) The image-based feature is a fixed length representation which makes it suitable to be combined with the framework of fingerprint encryption algorithm. So, the research on imaged-based feature is still active. Ordinal feature of this study is a image-based feature, with the above three advantages. Ordinal feature is intrinsic feature of fingerprint patterns and largely invariant to illumination changes. Moreover, compactness and low computational complexity of ordinal feature enable highly efficient fingerprint recognition. So it achieves a good trade-off between distinctiveness and robustness.This thesis focuses on application of ordinal feature for fingerprint identification and cancelable templates. Firstly, we propose using ordinal measure for fingerprint feature representation with the objective of characterizing qualitative relationship between fingerprint regions. Secondly, we fuse ordinal feature and fingerprint orientation estimated by Legendre polynomials to improve the verification performance. The algorithm was tested and compared with other image-based algorithm on FVC2002. Experiment results showed that the accuracy of our approach is comparable to the traditional LBP feature and storage space and computing complexity has been reduced. Finally, we proposed a cancelable scheme based on ordinal feature, which cancel the comprised template by control the parameters of multilobe differential filters(MLDFs) with random seed and reissues new template. Experiment on FVC2002 showed the scheme meet requirements of cancelable transform on matching performance, distinctiveness, noninvertibility and security.