QR steganographic image transmission system based on multimode fiber and deep learning

• DOI: 10.1364/ao.544443
• Publication Date: 2025-03-03T19:00:21Z
• Authors: Houru Zhao; YiQiang Zhang; Leihong Zhang; Dawei Zhang

In this paper, we propose a QR steganographic image transmission system based on a multimode optical fiber with deep learning, aiming to solve the image distortion problem due to mode dispersion in multimode optical fiber transmission and to improve the security and reliability of transmission. The system adopts the PIES-Net model to generate visually imperceptible steganographic images by embedding secret images in equal proportions into camouflaged images. Subsequently, the steganographic image is converted into a QR code, which utilizes its error correction capability to ensure that the original data can be recovered through redundant information, even if some of the information is lost or corrupted during transmission. After the QR code is transmitted over a multimode optical fiber, a scattered image is formed at the receiving end. In this paper, an improved SFNet model based on U-Net architecture is proposed for reconstructing QR codes and recovering the original information from the scattered image. The experimental results show that the system generates a covertly written image with high steganography, and the extracted secret image excels in visual quality, peak signal-to-noise ratio, and image correlation, and demonstrates excellent robustness and security in a variety of noise environments. This study provides what we believe are new ideas and important support for the development of multimode fiber communication and image steganography.