A high-capacity reversible data hiding based on improved rhombus prediction

Reversible data hiding algorithms based on prediction error histogram of rhombus prediction need excellent prediction performance to achieve more embedding capacity. However, the cost of improving the prediction accuracy is to reduce the number of prediction pixels, which results in a reduction in embedding capacity. It means high prediction accuracy and large embedding capacity become two contradictory conditions. In this paper, we proposed a reversible data hiding algorithm based on improved rhombus prediction, which can break through the upper limit of prediction pixel numbers and increase the embedding capacity. The proposed algorithm utilizes the characteristics of reversible data hiding process to carry out the second stage reversible embedding based on the traditional rhombus prediction. It can embed the additional data into the context pixels of the first stage rhombus prediction so that the embedding capacity will be increased significantly. At the same time, reusing the reversible data hiding characteristics can ensure the lossless recovery of the carrier image and secret information. Experiments show that the algorithm can improve embedding capacity significantly and ensure the high visual performance of marked images.