テラヘルツ（THz）リアルタイム高解像度イメージングを用いた導電性フィルム上の暗号化および復号化アプリケーション

In this work, we present a terahertz (THz) real-time quasi-near-field imaging system for low-cost and highly concealable information encryption and decryption. By exploiting the distinct THz transmission properties of graphite pencils with different compositions, we designed three types of encryption structures, including text, coating, and QR code patterns. All structures were successfully decrypted, demonstrating the feasibility of material-based THz encryption. However, the imaging clarity significantly deteriorates under ultra-low signal-to-noise ratio (SNR) conditions, which is a major challenge in practical applications. To address this issue, we propose a modified U-Net denoising method that integrates multi-frequency feature fusion and THz-specific physical priors. The network effectively suppresses noise while preserving fine structural details. Experimental results show that, compared with the conventional block-matching and 3D filtering (BM3D) algorithm, our method achieves superior performance, with peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and feature similarity (FSIM) values significantly improved. More importantly, encrypted information remains clearly recognizable even in complex noise environments. This work combines material-based encryption strategies, advanced THz imaging, and deep learning denoising into a unified framework. Owing to its low implementation cost, strong concealment, and enhanced stability, the proposed approach offers a promising pathway for extending THz imaging to secure communication, anti-counterfeiting, and information protection.