Fingerprint features extraction using curve-scanned DCT coefficients

Abstract

This paper proposes a fingerprint features extraction method using curve-scanned DCT coefficients. Generally, the fingerprint features contained in a fingerprint image, such as ridge line patterns and minutiae point, can be extracted from the DCT domain, and used for fingerprint matching. By considering the oscillate pattern contained in the top-left corner of DCT coefficients, we can divide those coefficients in curve-scanned fashion, extract DCT features from the divided DCT coefficients, and use them for matching purpose. To evaluate the proposed method, we use the k-NN classifier to measure the recognition rate, and compare the results obtained from our extraction method to that from the DWT based method [5] and the zigzag-scanned based method [6]. According to the results, the proposed method outperforms the other two, particularly in both recognition rate and processing time.