Fidelity Enhancement of Reversible Data Hiding for Images with Prediction-Based Concepts

Reversible data hiding has attracted more and more attention in recent years. With reversible data hiding, it requires embedding secret data into original image with devised algorithm at the encoder, and marked image can then be delivered to the decoder. At the decoder, both the secret data and original image should be perfectly separated from marked image to keep the reversibility. There are several practical ways to make reversible data hiding possible, and one of the latest methods belongs to the prediction-based method. By carefully manipulating differences between predicted and original images, reversible data hiding can be achieved. We propose an enhanced method for manipulating the difference histogram, and we observe the better performances than existing scheme in literature. Possible ways for enhancing embedding capacity are also pointed out for the extension of our method in the future.