| Biometric system deals with recognizing a person by determining the authenticity based on physiological and behavioral characteristics. Fingerprints are well known for their distinctiveness among individuals and have been widely used for personal authentication and recognition. This thesis studies the enhancement problem for fingerprint images with marginal quality due to a number of reasons (e.g., scar, moisture, dryness, over-pressure).; Our contribution is two fold. In single-copy enhancement, we have developed an image inpainting based approach to repair low-quality regions in a fingerprint images that are detected by ridge coherence. In multi-copy enhancement, we propose to mosaic multiple copies of the same finger by optical-flow based registration and Least-Square based interpolation. Experimental results are reported to validate the effectiveness of the proposed techniques. Some conclusions about the future study are made. |
