Intelligent Watermarking of Document Images as a Dynamic Optimization Problem

Intelligent watermarking (IW) allows adapting embedding parameter for each image and set of attacks using evolutionary computing. However, IW is not practical in real-world applications because of the computational cost of (evolutionary computing) algorithms that must be applied to optimize the parameters for each document image. It is however possible to formulate IW as a dynamic optimization problem (DOP), as similar images should result in similar fitness landscapes. Such formulation could allow a decrease in the computational burden of IW since dynamic optimization uses knowledge obtained in previous image optimizations in order to reduce the number of optimizations in some cases, replacing re-optimization by recalls to previously seen solutions. In this paper, a bi-tonal intelligent watermarking system is proposed and employed as a tool to characterize IW as a DOP. Proof of concept simulations using such approach resulted in a decrease of 93.5% in the computational cost with little impact on accuracy.