Abstract:Robust 3D mesh watermarking is a traditional research topic in computer graphics, which provides an efficient solution to the copyright protection for 3D meshes. Traditionally, researchers need manually design watermarking algorithms to achieve sufficient robustness for the actual application scenarios. In this paper, we propose the first deep learning-based 3D mesh watermarking network, which can provide a more general framework for this problem. In detail, we propose an end-to-end network, consisting of a wa… Show more
“…In PSNR with attacks, 14.7 KB image size is considered to evaluate the experiments. The average PSNR of proposed NCDG-LV for 10 models is 55.02, whereas the PSNR of existing 5 , 18 , 23 is 52.52, and 50.38, and 53.35 respectively.…”
Section: Resultsmentioning
confidence: 96%
“…9 . Despite the very high number of 3D mesh models for the simulation, the proposed method achieved a smaller distortion rate compared to Laplace–Beltrami 3D 18 , 3D-MDAQIM 5 , and Deep 3D mesh watermarking network 23 . Figure 9 Graphical representation of distortion rate.…”
Section: Resultsmentioning
confidence: 99%
“…The salient point detection time complexity is significantly smaller for a single 3D mesh model and higher for large number of models. For the first simulation run, the salient point detection time obtained by the proposed NCDG-LV method, Laplace–Beltrami 3D 18 , 3D-MDAQIM 5 , and Deep 3D mesh watermarking network 23 was determined to be 7.75, 9.25, 10.25 and 8.15, respectively. The better salient point detection time of NCDG-LV, which was 30%, 48% and 18% faster than the three other respective methods, can be attributed to the Nearest Centroid function employed via the Discrete Gaussian Kernel function.…”
Section: Resultsmentioning
confidence: 99%
“…Additionally, an attack layer perturbs the watermarked vertices to augment robustness against cropping, noise, rotation, translation, and scaling attacks. Wang et al 23 introduces the deep 3D mesh watermarking network, where the curvature consistency loss function is created to limit the local geometry smoothness of watermarked meshes in order to maintain the visual quality of 3D meshes. The architecture includes embedding, extracting sub-networks, and attack layers, employing topology-agnostic graph convolutions for flexible mesh handling.…”
Watermarking is one of the crucial techniques in the domain of information security, preventing the exploitation of 3D Mesh models in the era of Internet. In 3D Mesh watermark embedding, moderately perturbing the vertices is commonly required to retain them in certain pre-arranged relationship with their neighboring vertices. This paper proposes a novel watermarking authentication method, called Nearest Centroid Discrete Gaussian and Levenberg–Marquardt (NCDG–LV), for distortion detection and recovery using salient point detection. In this method, the salient points are selected using the Nearest Centroid and Discrete Gaussian Geometric (NC–DGG) salient point detection model. Map segmentation is applied to the 3D Mesh model to segment into distinct sub regions according to the selected salient points. Finally, the watermark is embedded by employing the Multi-function Barycenter into each spatially selected and segmented region. In the extraction process, the embedded 3D Mesh image is extracted from each re-segmented region by means of Levenberg–Marquardt Deep Neural Network Watermark Extraction. In the authentication stage, watermark bits are extracted by analyzing the geometry via Levenberg–Marquardt back-propagation. Based on a performance evaluation, the proposed method exhibits high imperceptibility and tolerance against attacks, such as smoothing, cropping, translation, and rotation. The experimental results further demonstrate that the proposed method is superior in terms of salient point detection time, distortion rate, true positive rate, peak signal to noise ratio, bit error rate, and root mean square error compared to the state-of-the-art methods.
“…In PSNR with attacks, 14.7 KB image size is considered to evaluate the experiments. The average PSNR of proposed NCDG-LV for 10 models is 55.02, whereas the PSNR of existing 5 , 18 , 23 is 52.52, and 50.38, and 53.35 respectively.…”
Section: Resultsmentioning
confidence: 96%
“…9 . Despite the very high number of 3D mesh models for the simulation, the proposed method achieved a smaller distortion rate compared to Laplace–Beltrami 3D 18 , 3D-MDAQIM 5 , and Deep 3D mesh watermarking network 23 . Figure 9 Graphical representation of distortion rate.…”
Section: Resultsmentioning
confidence: 99%
“…The salient point detection time complexity is significantly smaller for a single 3D mesh model and higher for large number of models. For the first simulation run, the salient point detection time obtained by the proposed NCDG-LV method, Laplace–Beltrami 3D 18 , 3D-MDAQIM 5 , and Deep 3D mesh watermarking network 23 was determined to be 7.75, 9.25, 10.25 and 8.15, respectively. The better salient point detection time of NCDG-LV, which was 30%, 48% and 18% faster than the three other respective methods, can be attributed to the Nearest Centroid function employed via the Discrete Gaussian Kernel function.…”
Section: Resultsmentioning
confidence: 99%
“…Additionally, an attack layer perturbs the watermarked vertices to augment robustness against cropping, noise, rotation, translation, and scaling attacks. Wang et al 23 introduces the deep 3D mesh watermarking network, where the curvature consistency loss function is created to limit the local geometry smoothness of watermarked meshes in order to maintain the visual quality of 3D meshes. The architecture includes embedding, extracting sub-networks, and attack layers, employing topology-agnostic graph convolutions for flexible mesh handling.…”
Watermarking is one of the crucial techniques in the domain of information security, preventing the exploitation of 3D Mesh models in the era of Internet. In 3D Mesh watermark embedding, moderately perturbing the vertices is commonly required to retain them in certain pre-arranged relationship with their neighboring vertices. This paper proposes a novel watermarking authentication method, called Nearest Centroid Discrete Gaussian and Levenberg–Marquardt (NCDG–LV), for distortion detection and recovery using salient point detection. In this method, the salient points are selected using the Nearest Centroid and Discrete Gaussian Geometric (NC–DGG) salient point detection model. Map segmentation is applied to the 3D Mesh model to segment into distinct sub regions according to the selected salient points. Finally, the watermark is embedded by employing the Multi-function Barycenter into each spatially selected and segmented region. In the extraction process, the embedded 3D Mesh image is extracted from each re-segmented region by means of Levenberg–Marquardt Deep Neural Network Watermark Extraction. In the authentication stage, watermark bits are extracted by analyzing the geometry via Levenberg–Marquardt back-propagation. Based on a performance evaluation, the proposed method exhibits high imperceptibility and tolerance against attacks, such as smoothing, cropping, translation, and rotation. The experimental results further demonstrate that the proposed method is superior in terms of salient point detection time, distortion rate, true positive rate, peak signal to noise ratio, bit error rate, and root mean square error compared to the state-of-the-art methods.
“…However, traditional 3D digital watermarking methods are relatively cumbersome, and researchers have also applied deep learning techniques to 3D digital watermarking. Wang et al (2022) proposed the first deep learning based 3D mesh watermarking network, providing a more universal 3D mesh watermarking framework. Yoo et al (2022) achieved embedding messages in 3D meshes and extracting them from 2D rendering.…”
The Neural Radiation Field (NeRF) has achieved significant results in the field of computer vision. However, training NeRF models requires a large amount of computing resources and training data. Once the model is stolen or abused by illegal users, it may cause huge losses to copyright owners. This article proposes a watermarking framework for copyright protection using Implicit Neural Representation (INR). After implicitly representing the watermark information, we embed the network representing the watermark information into the network of the NeRF model by designing a specific key. The copyright verifier extracts the watermark from the NeRF model using the key to prove the copyright of the NeRF model. To our knowledge, this is the first time INR has been introduced into the copyright protection of the NeRF model. The experimental results show that our method not only has high robustness and maintains good 3D reconstruction quality, but also achieves lossless (100%) extraction of watermark information, effectively protecting the copyright of the NeRF model.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
