A Parallel PSO Algorithm for a Watermarking Application on a GPU

In this paper, a research about the usability, advantages and disadvantages of using Compute Unified Device Architecture (CUDA) is presented, implementing an algorithm based on populations called Particle Swarm Optimization (PSO) (5). In order to test the performance of the proposed algorithm, a hide watermark image application is put into practice. The PSO is used to optimize the positions where a watermark has to be inserted. This application uses the insertion/extraction algorithm proposed by Shieh et al. (1). This algorithm was implemented for both sequential and CUDA architectures. The fitness function—used in the optimization algorithm—has two objectives: fidelity and robustness. The measurement of fidelity and robustness is computed using Mean Squared Error (MSE) and Normalized Correlation (NC), respectively; these functions are evaluated using Pareto dominance.