Extraction of Smooth and Thin Ridgelines from Fingerprint Images Using Geometric Prediction

Ridgelines play a vital role in a modern fingerprint identification system. However, due to inherent limitations in a fingerprint acquisition device, detection of ridgelines from an acquired image becomes a challenging task.Hence, several methodologies have been suggested over the years to detect ridgelines in a gray-scale fingerprint image, most of which produce thick ridgelines requiring subsequent thinning and smoothing in order to obtain the thinned ridge topography. Proposed in this paper is an efficient algorithm of extracting ridgelines from gray-scale fingerprint image using geometric prediction based on an efficient carry-forward mechanism of the knowledge acquired while traversing along the concerned ridgeline. Such a geometric tracing finally produces a ridge segment that is simultaneously smooth and thin, wherein lies the novelty of the algorithm.The algorithm is robust, efficient, and has minimal dependence on shareholding. On testing with benchmark databases, it is found to produce the desired output in terms of both runtime and quality.