Robust Image Watermarking Based on Texture Guidance and Dynamic Multi-Spectrum Attention

Tracking image sources and verifying copyright information is crucial in digital media communication. Digital image watermarking technology, widely used for copyright protection and source tracking, faces challenges in balancing imperceptibility and robustness under a certain embedded capacity. To address this problem, this paper proposes a robust image watermarking framework based on a deep invertible network, named DRIN. Based on the frequency-domain characteristics of robust steganography, we parallelly introduce a texture guidance module (GLCM) and a dynamic fine-grained DCT representation-based channel attention module (DFscaNet) into the network. These modules guide the model to effectively locate the texture features where high-frequency energy is concentrated in the cover image, improving the quality of the watermarked image and enhancing the robustness of the watermark. Experiments have demonstrated that the proposed method has superior performance, higher visual quality, and stronger robustness compared to the state-of-the-art methods.