Transformer-based Brightness Correction for Finger Vein Images

Weijian Luo,Xiong Zhang

DOI: https://doi.org/10.1117/12.3006040

2023-01-01

Abstract:Finger vein authentication, as one of the identity authentication technologies, has the advantages of high stability, strong concealment, and the ability to perform liveness detection. However, the performance of finger vein authentication is susceptible to variations in lighting conditions across different scenes. Finger vein images with low or high brightness can lead to the loss of vein texture, resulting in authentication failures. To extend the application of finger vein authentication models with excellent performance in normal lighting conditions to diverse lighting scenarios, this paper proposes a Transformer-based finger vein image brightness correction network. Leveraging the global interaction capability of self-attention, the proposed network achieves comprehensive control over the brightness information of images and effectively adapts the input image's brightness to the target brightness distribution. Extensive experiments are conducted to validate the effectiveness of the brightness correction network. The experimental results confirm that the proposed approach can effectively correct the brightness of finger vein images, thereby improving the accuracy and stability of the authentication system across different lighting conditions.