An Investigation on the Development of Digital Watermarking Technology for Anti-Screening Photography
TL;DR
In deep learning-based anti-screen capture digital watermarking technology, the paper focuses on the breakthroughs in enhancing robustness and security through auxiliary feature enhancement, end-to-end joint training, and adversarial training under deep learning.
Abstract
With the widespread dissemination of captured images and digital images, the leakage of information through photos taken on smart terminals faces severe security challenges. This paper systematically reviews the research progress of anti-screen capture watermarking technology in recent years, dividing representative anti-screen capture digital watermarking technologies into two main categories: those based on spatial domain embedding and those based on deep learning. Within these technologies, based on the different embedding domains of the watermark information, anti-screen capture watermarking is further divided into spatial domain embedding and frequency domain embedding. The paper summarizes and categorizes anti-screen capture watermarking technologies from the perspective of the evolution of "feature point robustness." In deep learning-based anti-screen capture digital watermarking technology, the paper focuses on the breakthroughs in enhancing robustness and security through auxiliary feature enhancement, end-to-end joint training, and adversarial training under deep learning. Finally, it points out that self-synchronization, high visual fidelity, and model security are key directions for future research. This review helps readers gain a deeper understanding of the history and advanced technologies of digital watermarking in the anti-screen capture field, aiming to provide a reference for cross-media information security research.