Deep Boosting Robustness of DNN-based Image Watermarking via DBMark
This work addresses the need for more robust digital watermarking techniques for applications like copyright protection, though it appears incremental as it builds on existing DNN-based approaches.
The paper tackles the problem of improving robustness in DNN-based image watermarking by proposing DBMark, a framework that synergizes invertible neural networks and effective watermark feature generation, achieving superior performance under various distortions like dropout, cropout, and JPEG compression compared to state-of-the-art methods.
Image watermarking is a technique for hiding information into images that can withstand distortions while requiring the encoded image to be perceptually identical to the original image. Recent work based on deep neural networks (DNN) has achieved impressive progression in digital watermarking. Higher robustness under various distortions is the eternal pursuit of digital image watermarking approaches. In this paper, we propose DBMARK, a novel end-to-end digital image watermarking framework to deep boost the robustness of DNN-based image watermarking. The key novelty is the synergy of invertible neural networks (INN) and effective watermark features generation. The framework generates watermark features with redundancy and error correction ability through the effective neural network based message processor, synergized with the powerful information embedding and extraction abilities of INN to achieve higher robustness and invisibility. The powerful learning ability of neural networks enables the message processor to adapt to various distortions. In addition, we propose to embed the watermark information in the discrete wavelet transform (DWT) domain and design low-low (LL) sub-band loss to enhance invisibility. Extensive experiment results demonstrate the superiority of the proposed framework compared with the state-of-the-art ones under various distortions such as dropout, cropout, crop, Gaussian filter, and JPEG compression.