Cross-Area Finger Vein Recognition via Hierarchical Sparse Representation

The prior studies in finger vein recognition have mainly focused on personal identification based on images with same area. However, the upgrade of finger vein acquisition devices is inevitable, and therefore the scale variation of acquisition windows among various devices may cause the cross-area finger vein recognition problem. To address this problem, a hierarchical sparse representation-based cross-area finger vein recognition method is proposed in this paper. In the proposed method, the first layer locates the potential corresponding regions in each full training image for the small-area testing image based on coding coefficients on the image-specific dictionary, and the small-area testing image is classified in the second layer by its reconstruction error on the compact dictionary. In addition, the method is performed on the original and down-sampled images, and the weighted sum of the reconstruction errors on two kinds of images are used in recognition. The experiments are performed on two widely used finger vein databases, and the experimental results show that the proposed method achieves 91.35% and 78.94% recognition rates on cross-area finger vein recognition.