Robust Display-to-Camera Communication via Location Error-Tolerant Deep Data Embedding

This paper proposes a location error-tolerant data embedding technique for display-to-camera (D2C) communication systems. The method is designed to enable robust data transmission while maintaining high image fidelity, facilitating simultaneous digital content display and data communication. To address the common issue of alignment and localization inaccuracies in D2C systems, the proposed approach defines specific regions of interest to ensure robustness against object detection model location errors. The architecture employs an expanding and contracting network structure for the encoder to achieve seamless data integration, while the decoder utilizes a computationally efficient “thin” structure for rapid data extraction. To improve performance in diverse environments, various distortion models were integrated into the system training. The system’s effectiveness was evaluated by measuring the bit error rate under conditions of Gaussian noise, blur, simulated localization errors, and real-world distortions. Image quality was validated using peak signal-to-noise ratio and the structural similarity index measure. The results indicate that the proposed technique maintains superior image quality and achieves reliable data recovery even in the presence of significant localization errors. These findings suggest that the approach provides a stable and effective solution for practical mobile-based D2C communication.