Fingerprint feature extraction using block-direction on reconstructed images

Fingerprints have been used as unique identifiers of individuals for a very long time. As fingerprint databases are characterized by their large size and may contain noisy and distorted images, an efficient representation of the images is essential for a reliable identification. Considering fingerprints as sample images from non-stationary processes with flow patterns, we propose here a robust technique to extract their features. The unique properties of fingerprint textures are used to enhance the images and improve the fidelity of their features. The ridges are extracted from enhanced foreground areas based on local dominant ridge directions. The resulting bit-mapped images are thinned and smoothed to detect structural features. A large number of false features are eliminated in the proposed post-processing stage. The proposed algorithm results in an efficient and fast representation of fingerprints which accurately retains the fidelity in minutiae.