Accelerating Robust Watermarking through Parallelization

The rapid online usage of digital images has pushed us to accelerate the watermarking method to protect it. Lately, in order to accelerate the applied science and engineering methods multicore architectures are used more and more. Therefore, at this research we have tried to increase the speed of watermarking method on GPU by CUDA. This acceleration by CUDA allows rapid embedding and extraction of the watermarks. For the sake of the fact that better methods have more computational cost, parallelization leads up to the use of a better watermarking method which has a better performance in terms of imperceptibility, robustness, and security. It assists to the rapid secure use of digital images, as well. The results show that the speeds of CUDA implementations are superior to multi cores which have been implemented by OpenMP, although, when the images are big enough OpenMP implementations show a slightly better performance than sequential implementations. The proposed method is evaluated on an NVIDIA 940MX GPU which can reduce the execution time by 50% in comparison with sequential implementations.