Predictive reversible data hiding schemes for enhanced embedding capacity

Abstract

In reversible image steganographic schemes, an embedding algorithm implants the secret bits into a few high-frequency contents in its embedding space such that a de-embedding algorithm can extract the secret and reconstruct the original image. Most of these schemes increase the quantity of these embeddable contents through some pre-processing mechanisms like computing the transformed coefficients and measuring prediction errors by applying a prediction process. The prediction error based schemes are able to implant a higher number of bits, because the prediction errors are mostly distributed in zero or around the zero in the prediction error histogram. The embedding performance of these schemes depends on their ability to maximize the quantity of these high-frequency embeddable errors. Towards improving the frequency of these embeddable errors, this thesis aims in its first stage to improve the prediction accuracy of the existing multi-block centre reference based predictor by more rationally weighing the pixels in the prediction rules. The thesis also improves the frequency of embeddable errors by applying multiple predictors and computing the optimal error for each pixel from these multiple prediction errors or from a set of hybrid errors generated by applying these errors in a set of linear equations. The thesis demonstrates a policy of repeatedly implanting secret bits in the embeddable errors. The proposed methods of generating embeddable content for each pixel and of controlling the generation of the embeddable contents according to the demand of the application have further enriched the arena of data hiding technology. The thesis also contributes in enhancing the embedding capacity by applying prediction methodologies in the image distortion based reversible processes. It strengthens the security of the implanted data by encapsulating the implemented security levels. All the proposed schemes have the potential to make significant contributions towards hiding the large volume of data as well as to demonstrate the superior performances over their competing ones. The contribution of this thesis will hasten the arrival of new digital communication era in securing the transmission of copyrights, evidence, investigation reports, scientific results and political documents in the area of medical, forensic, law-enforcing agencies and military use.