Dense D2C-Net: dense connection network for display-to-camera communications.
Optics Express
|September 15, 2023
Summary
This study presents Dense D2C-Net, a novel display-to-camera (D2C) method using deep convolutional neural networks (DCNNs) to embed data in images. The scheme achieves superior performance in real-world tests, maintaining visual quality and data integrity.
Area of Science:
- Computer Science
- Electrical Engineering
- Information Technology
Background:
- Display-to-camera (D2C) communication enables data transfer via visual content.
- Existing DCNN-based D2C schemes face challenges in visual quality and robustness.
Purpose of the Study:
- Introduce Dense D2C-Net, a novel DCNN-based scheme for unobtrusive data embedding and extraction via visual content.
- Enhance data integrity and visual quality in D2C communication.
Main Methods:
- Developed Dense D2C-Net, a DCNN architecture with inter-layer connections and feature reuse for encoding.
- Utilized the Y channel for embedding binary data due to its robustness.
- Integrated hybrid layers and noise layers to improve data hiding and mitigate channel distortions.
- Employed 2D convolutional layers for data extraction at the decoder.
Main Results:
- Demonstrated superior performance of Dense D2C-Net compared to conventional DCNN-based D2C schemes.
- Validated effectiveness across various parameters: transmission distance, capture angle, display brightness, and camera resolution.
- Maintained high visual quality of the cover image with embedded data.
Conclusions:
- Dense D2C-Net offers a robust and efficient solution for display-to-camera communication.
- The proposed scheme provides a significant advancement in embedding and extracting data through visual content.
- Effective for real-world applications utilizing smartphone cameras and digital displays.
