Intelligent Watermarking with Multi-objective Population Based Incremental Learning

Intelligent watermarking techniques use nonconventional methods like Evolutionary Computation techniques to satisfy the trade-off between integrity and authenticity of digitized documents. In this paper, we propose a multi-objective Population Based Incremental Learning module for an intelligent watermarking system for grayscale images to optimize embedding watermarks that satisfy the trade-off between quality and robustness. The multi-objective formulation provides set of non-dominated solutions rather than single solution, which allows tuning the quality and robustness for several attacks, without the need for an expensive re-optimization process due to changing the priority of different objectives after the optimization process. Different formulations for the optimization problem were investigated to achieve best fitness for different objectives. The best fitness achieved for all objectives are compared for different formulations. Simulation results indicate better fitness for different objectives with multi-objective formulation, and faster convergence using incremental learning techniques.