DEEP REPRESENTATION BASED FEATURE EXTRACTION AND RECOVERING FOR FINGER-VEIN VERIFICATION

Abstract

Finger-vein biometrics has been extensively investigated for personal verification. Despite recent advances in fingervein verification, current solutions completely depend on domain knowledge and still lack the robustness to extract finger-vein features from raw images. This paper proposes a deep learning model to extract and recover vein features using limited a priori knowledge. Firstly, based on a combination of known state of the art handcrafted finger-vein image segmentation techniques, we automatically identify two regions: a clear region with high separability between finger-vein patterns and background, and an ambiguous region with low separability between them. The first is associated with pixels on which all the segmentation techniques above assign the same segmentation label (either foreground or background), while the second corresponds to all the remaining pixels. This scheme is used to automatically discard the ambiguous region and to label the pixels of the clear region as foreground or background. A training dataset is constructed based on the patches centered on the labeled pixels. Secondly, a Convolutional Neural Network (CNN) is trained on the resulting dataset to predict the probability of each pixel of being foreground (i.e. vein pixel) given a patch centered on it. The CNN learns what a fingervein pattern is by learning the difference between vein patterns and background ones. The pixels in any region of a test image can then be classified effectively. Thirdly, we propose another new and original contribution by developing and investigating a Fully Convolutional Network (FCN) to recover missing fingervein patterns in the segmented image. The experimental results on two public finger-vein databases show a significant improvement in terms of finger-vein verification accuracy.

Existing System

Gabor wavelets, Random Forest Method

Proposed System

We propose a deep learning model for finger-vein verification. Our approach aims, first, at segmenting foreground (vein) pixels from background pixels by predicting the probability of a pixel to belong to a vein pattern given limited knowledge, and, second, at recovering missing vein patterns. Compared to current state of the art segmentation and recovering approaches, that are based on image processing techniques, our approach does not segment or recover an image based only on its pixels and their correlations, but it does so by relying also on rich statistics on nonlinear pixel correlations, through a hierarchical feature representation learned by a deep neural network from a large training set. This is a major advantage over traditional approaches as relying only on noisy input images for segmentation or vein recovery may lead to severe errors.

Architecture

BLOCK DIAGRAM