An adaptive method for vein recognition enhancement using deep learning

Authentication methods based on some human traits including fingerprints, face, and palm prints, have developed significantly, and currently are sophisticated enough to be reliably considered for identification purposes. As state-of-the-art techniques, computational methods have recently been introduced to extract vein patterns for identification purposes. However, these methods suffer in a lack of image enhancement which the vein pattern cannot be uncovered properly. To address image enhancement weaknesses, this paper proposes a new image enhancement method based on deep learning, which can improve the robustness of vein recognition and feature extraction. Making use of deep neural network techniques, a sparse auto-encoder algorithm is applied to develop an adaptive method for image enhancement. We use a pair of synchronised colour and near-infrared images for verifying the outcome of the proposed method and measure the length of veins. The experimental results on 605 colour images and 605 infrared images from 66 forearms, 66 thighs, 33 chests, 33 ankles, and 50 random internet images show that the proposed enhancement method significantly improves the feature extraction procedure.