Automatic Feature Extraction from non-uniform Finger Vein Image and its Application to Personal Identification

A new method to extract vein patterns from an unclear finger image is proposed. Finger vein pattern can be used for personal identification technology. The finger vein image that is captured with infrared light contains not only the vein patterns but also irregular shades and noises. They result from the different thickness between finger bones and muscles and, light intensity fluctuations also produce them. Therefore, the robustness and the tolerance to the uneven luminance and noises in pattern extraction algorithm are required. In this paper, we propose a method for extracting the finger vein patterns in the image that is unclear and/or affected by the fluctuations of LED intensity. The method is based on special line tracking that starts at various positions. The robustness of the finger vein extraction method against LED intensity fluctuation is tested and the result shows the performance of the method is more robust than the conventional method. In the evaluation of the personal identification using the proposed extraction method, the finger vein patterns of 678 people are collected. The performance is 0.145% equal error rate in the evaluation.