Multiscale Feature Extraction of Finger-Vein Patterns Based on Curvelets and Local Interconnection Structure Neural Network

In this paper, we originally propose a multiscale feature extraction method of finger-vein patterns based on curvelets and local interconnection structure neural networks. The curvelets is used to perform the multiscale self-adaptive enhancement transform on the finger-vein image and a neural network with local interconnection structure is designed to extract the features of the finger-vein pattern. This method has the following features: firstly, the feature of finger-vein is line feature, or anisotropy, which is more suitable to be processed by curvelets than wavelets, especially when dealing with the obscure anisotropic features. Secondly, when the multiscale self-adaptive enhancement transform is applied to the finger-vein image, the finger-vein pattern is emphasized and noises are refrained greatly. Thirdly, a local interconnection neural network with linear receptive field is designed to deal with finger-vein patterns of different thickness and capture the patterns. Fourthly, the method is very fast by using the integral image method. The experimental results show the proposed method is superior to other methods in finger-vein feature extraction and solve the problem of how to extract features from obscure images efficiently. The EER of the proposed method is 0.128%