Currently, most transactions and exchanges are conducted through the Internet thanks to technological tools, running the risk of the falsification and distortion of information. This is due to the massive demand for the virtual world and its easy access to anyone. Image watermarking has recently emerged as one of the most important areas for protecting content and enhancing durability and resistance to these kinds of attacks. However, there is currently no integrated technology able to repel all possible kinds of attacks; the main objective of each technology remains limited to specific types of applications, meaning there are multiple opportunities to contribute to the development of this field. Recently, the image watermarking field has gained significant benefits from the sudden popularity of deep learning and its outstanding success in the field of information security. Thus, in this article, we will describe the bridge by which the watermarking field has evolved from traditional technology to intelligent technologies based on deep learning.

Nowadays, with the development of technology and the popularity of smartphones, dealing with captured documents has become an inevitable necessity. The print-cam process produces serious attacks during capturing images, especially perspective distortions. In this paper, we have improved the Fourier watermarking method proposed by Gourrame et al. [1] for Print-Cam attacks which is based on a specific correction preprocess. This correction consists of a frame-based perspective rectification for the captured images. Effectively, Instead of applying a classical Hough detecting-based technique to restore the geometric distortions, we have implemented a new deep learning neural architecture that enables us to segment the image more accurately. According to the tests carried out, this architecture allowed us to improve the robustness of the global watermarking method.

Image watermarking often involves the use of handheld devices under non-structured conditions for authentication purposes, particularly in the print-cam process where smartphone cameras are used to capture watermarked printed images. However, these images frequently suffer from perspective distortions, making them unsuitable for automated information detection. To address this issue, Cam-Unet, an end-to-end neural network architecture, is presented to predict the mapping from distorted images to rectified ones, specifically tailored for print-cam challenges applied to ID images. Given the limited availability of large-scale real datasets containing ground truth distortions, we created an extensive synthetic dataset by subjecting undistorted images to print-cam attacks. The proposed network is trained on this dataset, using various data augmentation techniques to improve its generalization capabilities. Accordingly, this paper presents an image watermarking system for the print-cam process. The approach combines Fourier transform-based watermarking with Cam-Unet as perspective distortion correction. Results show that the proposed method outperforms existing watermarking approaches typically employed to counter print-cam attacks and achieves an optimal balance between efficiency and cost-effectiveness.

Robust image watermarking that can resist camera shooting has gained considerable attention in recent years due to the need to protect sensitive printed information from being captured and reproduced without authorization. Indeed, the evolution of smartphones has made identity watermarking a feasible and convenient process. However, this process also introduces challenges like perspective distortions, which can significantly impair the effectiveness of watermark detection on freehandedly digitized images. To meet this challenge, ResNet50-based ensemble of randomized neural networks (ReRNet), a recursive convolutional neural network-based correction method, is presented for the print-cam process, specifically applied to identity images. Therefore, this paper proposes an improved Fourier watermarking method based on ReRNet to rectify perspective distortions. Experimental results validate the robustness of the enhanced scheme and demonstrate its superiority over existing methods, especially in handling perspective distortions encountered in the print-cam process.