| The technique presented here is a fast, feature-based one to classify fingerprint images into five general groups. The uniqueness of this research is in combining the adaptive singular point detection algorithm with the coarse classification one. The first step in the technique enhances the raw image and finds the directional image. The core and delta correlation functions are then found. An adaptive threshold value is used to detect the singular points (delta and core points) from the correlation functions. Depending on the position, the fingerprints are classified into the Wirbel class (whorl and twin loop) or the Lasso class (arch, tented arch, right loop, or left loop), which is a coarse, rule-based classification method. The advantage of this algorithm is in using the adaptive threshold over the fixed threshold. The classification of the fingerprints into five general groups narrows down the search space in large fingerprint databases. The use of directional masks also speeds up the process. |
