Abstract

Visible watermark removal which involves watermark cleaning and background content restoration is pivotal to evaluate the resilience of watermarks. Existing deep neural network (DNN)-based models still struggle with large-area watermarks and are overly dependent on the quality of watermark mask prediction. To overcome these challenges, we introduce a novel feature adapting framework that leverages the representation modeling capacity of a pre-trained image inpainting model. Our approach bridges the knowledge gap between image inpainting and watermark removal by fusing information of the residual background content beneath watermarks into the inpainting backbone model. We establish a dual-branch system to capture and embed features from the residual background content, which are merged into intermediate features of the inpainting backbone model via gated feature fusion modules. Moreover, for relieving the dependence on high-quality watermark masks, we introduce a new training paradigm by utilizing coarse watermark masks to guide the inference process. This contributes to a visible image removal model which is insensitive to the quality of watermark mask during testing. Extensive experiments on both a large-scale synthesized dataset and a real-world dataset demonstrate that our approach significantly outperforms existing state-of-the-art methods. The source code is available in the supplementary materials.

Framework

Experiment

Conclusion

In conclusion, this paper introduces an innovative feature adapting framework tailored for the challenging task of large-area visible watermark removal. The proposed framework leverages specialized components, including a watermark component cleaning branch and a background content embedding branch, both equipped with transposed attention modules for enhanced feature extraction. The integration of gated fusion modules further refnes the image inpainting backbone, facilitating accurate reconstruction of watermarked regions by incorporating prompt information within the extracted features. Additionally, the model exhibits adaptability to imprecise watermark masks through the incorporation of a coarse segmentation mask. Empirical evaluations conducted on two datasets demonstrate the effectiveness of our method, showcasing its state-of-the-art performance in comparison to various existing approaches.