An effective algorithm for detecting valley of fingerprint based on wavelet transform

Personal authentication systems (PAS) using fingerprints are produced commercially and many profical applications in the real world have been reported many examples. However, in applying PAS, (1) deterioration of image quality by the influence of oily dirt adhering on the finger surface and (2) psychological resistance to using an input device touched by many users pose a difficult problem. The simplest method for solving this problem is to perform a visual input without physical contact. However, distinction of ridges and valleys is difficult with such a visual input in addition, edge extraction using differentiation filters etc. to portray images is another issue that has to be overcome. Therefore, a technique for extracting the valleys of fingerprints from a noncontact visual input using wavelet transform is proposed in this study. This technique, involved wavelet transforms and zone division before performing histogram equalization to the luminosity ingredient of the input picture. An edge picture was obtained by deleting the low-frequency signals from the divided signal group. Moreover, noise was removed by performing 0 value quantization on the high-frequency signals. After performing binarization, taking the curved surface form of the fingertip to obtain the edge picture without noise, a good valley was extracted by performing thinning with the algorithm of Deutsch. Our study shows the validity of the proposed technique; a reliable fingertip picture with a noncontact visual input was established.