基于噪声感知引导的鲁棒特征嵌入图像水印方法
By constructing a noise-aware guided robust feature embedding image watermarking network, the problem of balancing high quality and robustness in traditional methods is solved, achieving stable extraction of watermark information in complex environments, which is suitable for copyright protection and content traceability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHANDONG NORMAL UNIV
- Filing Date
- 2026-04-27
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional image watermarking methods struggle to improve robustness against attacks while maintaining high quality. Deep learning-based methods generate poor-quality watermarked images at low bit error rates, and the watermark extraction is also less robust.
A robust feature embedding image watermarking network based on noise perception guidance is constructed. The parameters are directly learned through the noise layer and indirectly updated through the decoder. The discriminator is combined as an adversarial network to optimize the network parameters to ensure the similarity between the carrier image and the watermarked image. A robust attention mechanism and an adversarial loss function are adopted to generate high-quality watermarked images.
While generating high-quality watermarked images, it significantly improves the robustness against attacks and can stably extract watermark information in complex environments, making it suitable for copyright protection and content traceability.
Smart Images
Figure CN122089551B_ABST
Abstract
Citation Information
Patent Citations
Image watermarking method based on space channel interactive attention mechanism
CN121582049A