A hybrid approach for colour image watermarking: wiener inverse transform factorization and dilated random graph attention network with Osprey optimization
Umaeswari Pargunaraj, Hariharan N, Reena Roy R, B. Sarala · The Imaging Science Journal · 2025
This paper presents a robust and imperceptible image watermarking framework that ensures content authenticity and protects intellectual property. Most existing techniques struggle to balance robustness against attacks and imperceptibility. The proposed approach combines transform-domain embedding, adaptive attention-based extraction, and an Osprey-inspired optimization algorithm for watermark strength adjustment. Wiener-Hopf factorization preserves high-image quality, while a graph-attention decoder enhances resistance to compression, geometric distortions, and noise. Experimental results on benchmark datasets show superior performance with PSNR 73 dB, SSIM 0.99, NC 0.99, and BER 0.02, outperforming DARI-Mark, WHT, PSO-SRCNN, MFrRHFM, and AdvEWM. Embedding and extraction times of 0.08 s and 0.06 s demonstrate computational efficiency. The framework's high robustness and imperceptibility make it suitable for secure digital rights management and healthcare imaging, ensuring authenticity and integrity without compromising diagnostic quality.