Abstract:
Snapshot spectral imaging technology based on the principle of compressed sensing can introduce additional spectral information and solve the problem of inadequate perfor...View moreMetadata
Abstract:
Snapshot spectral imaging technology based on the principle of compressed sensing can introduce additional spectral information and solve the problem of inadequate performance of existing optical sensors in dealing with high-fidelity masks. However, using compressed sensing for spectral image reconstruction requires a large amount of computational resources and has limited accuracy, making it the greatest obstacle for the practical application of snapshot spectral imaging technology. In this paper, we propose a new paradigm for Face Anti-Spoofing (FAS) based on snapshot spectral imaging technology. Our method does not require the reconstruction of compressive sensing spectral images, but instead directly treats the snapshot spectral imaging hardware as an optical encoder and designs the end-to-end Snapshot Spectral Imaging Network (SSINet) to decode the optical encoding results to obtain FAS results, which constitutes a software-hardware integrated FAS method. Experimental results show that our method not only surpasses previous state-of-the-art methods based on RGB images, but also performs the better than conventional spectral reconstruction methods in both efficiency and accuracy, with a nearly thousand-fold increase in FAS speed and an Average Classification Error Rate (ACER) decrease of 12.29%.
Date of Conference: 30 June 2024 - 05 July 2024
Date Added to IEEE Xplore: 09 September 2024
ISBN Information: