Robust watermarking using diffusion of logo into auto-encoder feature maps

Abstract Digital content has grown dramatically in recent years, leading to increased attention to copyright. Image watermarking has been considered one of the most popular methods for copyright protection. With the recent advancements in the application of deep neural networks in image processing, these networks have also been used in image watermarking. Robustness and imperceptibility are two challenging features of watermarking methods that the trade-off between them should be satisfied. In this paper, we propose to use an end-to-end network for watermarking. We use a convolutional neural network (CNN) to control the embedding strength based on the images’ content. Dynamic embedding helps the network to have the lowest effect on the visual quality of the watermarked image. Different image processing attacks are simulated as a network layer to improve the robustness of the model. Our method is a blind watermarking approach that replicates the watermark string to create a matrix of the same size as the input image. This helps the network to spread the watermark information in a wider space. Therefore, by removing some parts of the image by attacks, the watermark information can be retrieved from other parts. Using dilation layers is another innovation of the proposed method that helps the watermark spread in the wider neighborhood. This will improve the robustness of the method against image processing attacks during extraction. Instead of diffusing the watermark data into the input image, we inject the data into the feature space and force the network to do this in regions that increase the robustness against various attacks. Having embedding, extraction, and attack layer in an end-to-end network helps increase robustness and imperceptibility. Experimental results show the proposed method’s superiority in imperceptibility and robustness compared to the state-of-the-art algorithms.