Watermark extraction using blind image separation and sparse representation

Abdeldjalil Aissa El Bey (1, 2) , Amina Saleem (3) , Karim Abed-Meraim (4) , Azzedine Beghdadi (3)

1 Lab-STICC_TB_CACS_COM
2 SC - Département Signal et Communications
3 L2TI - Laboratoire de Traitement et Transport de l'Information
4 ENST - Ecole Nationale Supérieure des Télécommunications

This paper proposes a new image watermarking technique, which employs an Iterative sparse blind separation algorithm (ISBS) and a new sparsity based BSS algorithm (New ISBS algorithm or NISBS) for watermark extraction. The ISBS algorithm performs the separation based on the optimization of an Lp norm based contrast function. The new sparsity based BSS algorithm employs a smooth approximation of the absolute value function as the cost function. The watermark embedding of the images is performed in the spatial domain. The results of the simulations demonstrate the validity and performance of these two techniques for watermark extraction.

Domaines: Traitement du signal et de l'image [eess.SP]

Abdeldjalil Aissa El Bey, Amina Saleem, Karim Abed-Meraim, Azzedine Beghdadi. Watermark extraction using blind image separation and sparse representation. WoSPA 2008 : 5th International Workshop on Signal Processing and its Applications, Mar 2008, Sharjah, United Arab Emirates. ⟨hal-01770985⟩