A Robust and Blind Image Watermarking System Based on Deep Neural Networks
Frank Y. Shih · 2025
Digital image watermarking refers to the process of hiding certain messages into cover images. Incorporating deep neural networks with image watermarking has attracted increasing attentions during recent years. However, the robustness issue remains a challenge in the watermarking schemes of applying deep neural networks. In this paper, we present a robust and blind image watermarking system using deep convolutional neural networks, where the watermark embedding and extraction rules are learned and generalized. The robustness is achieved without any prior knowledge of possible attacks and distortions. Experimental results confirm that the proposed system achieves higher capacity as well as higher robustness, as comparing against several state-of-the-art techniques. A challenging application of watermark extraction on camera-captured images is also presented to validate the practicality of the proposed system.