Tatouage Robuste Et Aveugle Dans Le Domaine Des Valeurs Singulieres

Digital watermarking consists on inserting a mark into an image to protect it against copies. The heaviness of the extraction procedure with the old methods urged us to look for a new algorithm in the singular values domain which would be blind : it does not require the original image to extract the mark. We propose a new robust method which consists on inserting the bits of the mark into the singular values matrix. Contrary to most of the watermarking algorithms, it is blind and the results show that our method is robust against the JPEG compression, the reduction of colors (GIF) and the histogram spreading. So, we were able to obtain PSNR = 49,63 dB.

💡 Research Summary

The paper addresses two persistent challenges in digital image watermarking: the need for the original (unwatermarked) image during extraction (non‑blind schemes) and the high computational cost of many existing methods. To overcome these issues, the authors propose a blind watermarking algorithm that embeds the watermark directly into the singular values of an image’s Singular Value Decomposition (SVD).

In the embedding phase, the original image is decomposed into three matrices U, Σ, and Vᵀ, where Σ contains the singular values on its diagonal. The binary watermark is quantized into a bit stream {b₁, b₂, …, b_k}. For each selected singular value Σ