Fingerprint feature enhancement using block-direction on reconstructed images

Fingerprints have been used as unique identifiers of individuals for a very long time. The identification of fingerprint (FP) images is based on matching the features of a query FP, against those stored in a database. As FP databases are characterized by their large size and may contain noisy and distorted query images, an efficient and robust representation of FP images is essential for a reliable identification. Assuming FPs to be sample images from non-stationary processes of textured images of flow patterns, we propose here a new technique for preprocessing FP images for the purpose of identification. In the proposed algorithm, enhancement as well as ridge extraction processes are based on local dominant ridge directions. The obtained thinned image is then smoothed using morphological operations to detect FP structural features. The proposed algorithm results in an efficient, robust and fast representation of FPs which accurately retains the fidelity in minutiae (ridge endings and bifurcations).