Method of parallel circuits for fingerprint image recognition

New method of fingerprint image processing is described. The method is based on simultaneous tracking of ridges and valleys directions, instead of tracking ridges. In areas of good quality, the directions can be determined confidently both for ridges and valleys. But in areas of poor quality, there is a problem with accurate measuring of ridge flow. Given a pixel on filtered fingerprint image, the chains of pixels of minimal and maximal intensity are tracked. Based on the results of tracking, the hypotheses on ridge flow direction for ridges and valleys are generated. Then, the results are fused in order to obtain estimation of ridge flow direction and density. It allows to significantly improve the accuracy of ridge flow direction and density in the areas of poor quality. Finally, it is used to form the field of flows and its confidential (quality) matrix. More accurate measuring of ridge direction improves fingerprint recognition by more accurate feature extraction: ridge endings and bifurcations.