Pre-processing of fingerprint images

Fingerprint recognition remains as one of the most prominent biometric identification methods. The quality of the fingerprint image is the most significant factor in a reliable matching process. Any preprocessing algorithm should aim to enhance the quality of the existing features without creating false features. In this paper we describe a regional average thresholding (RAT) scheme for 256 grey scale images. The algorithm thresholds by moving a small, fixed size window on 256 colour grey scale image and uses a novel averaging scheme. The thresholded image is then operated on by a median filtering (3×3 mask or 5×5 mask) algorithm which eliminates any random noise contained in a thresholded fingerprint image. The filtered image is finally operated on by a ridge following thinning algorithm. We compare the results of using this algorithm on median filtered and nonfiltered images