Reversible data hiding based on PDE predictor

In this paper, we propose a prediction-error expansion based reversible data hiding by using a new predictor based on partial differential equation (PDE). For a given pixel, PDE predictor uses the mean of its four nearest neighboring pixels as initial prediction, and then iteratively updates the prediction until the value goes stable. Specifically, for each pixel, by calculating the gradients of four directions, the direction with small magnitude of gradient will be weighted larger in the iteration process, and finally a more accurate prediction can be obtained. Since PDE predictor can better exploit image redundancy, the proposed method introduces less distortion for embedding the same payload. Experimental results show that our method outperforms some state-of-the-art methods.