Performance comparison of orientation sensitive filters for low quality fingerprints

摘要

The efficiency of Fingerprint Authentication System (FAS) highly depends on the image quality of the print. Fingerprints, particularly on crime spots are of very poor quality (latent). Existing FAS systems highly falter in cases of samples obtained from laborers, soldiers etc, working in harsh and tedious environments. There is a need of some fingerprint enhancement algorithms to reconstruct low quality fingerprint images. We thus formulate a comparative analysis of orientation sensitive filtering mechanisms evolved with traditional and modified gabor filters (TGF and MGF). The paper proposes to develop variation in the kernel as a function of the wavelength. Simulations performed on large low quality image database approve an improvement of 4.29% in FNMR (False Non-Match Rate) @FMR=0.1587% in case of MGF as compared to TGF.