Efficient Reversible Data Hiding Based on Connected Component Construction and Prediction Error Adjustment

Abstract

To achieve a good trade-off between the data-embedding payload and the data-embedding distortion, mainstream reversible data hiding (RDH) algorithms perform data embedding on a well-built prediction error histogram. This requires us to design a good predictor to determine the prediction errors of cover elements and find a good strategy to construct an ordered prediction error sequence to be embedded. However, many existing RDH algorithms use a fixed predictor throughout the prediction process, which does not take into account the statistical characteristics of local context. Moreover, during the construction of the prediction error sequence, these algorithms ignore the fact that adjacent cover elements may have the identical priority of data embedding. As a result, there is still room for improving the payload-distortion performance. Motivated by this insight, in this article, we propose a new content prediction and selection strategy for efficient RDH in digital images to provide better payload-distortion performance. The core idea is to construct multiple connected components for a given cover image so that the prediction errors of the cover pixels within a connected component are close to each other. Accordingly, the most suitable connected components can be preferentially used for data embedding. Moreover, the prediction errors of the cover pixels are adaptively adjusted according to their local context, allowing a relatively sharp prediction error histogram to be constructed. Experimental results validate that the proposed method is significantly superior to some advanced works regarding payload-distortion performance, demonstrating the practicality of our method.

An Ensemble Learning Approach for Reversible Data Hiding in Encrypted Images with Fibonacci Transform

“…RDH schemes based on the difference expansion technique [41] embed secret data into the differences between the neighboring pixel values. RDH schemes are also explored for use with different image datasets, such as natural images [42], medical images [43], etc. Though many RDH schemes have been described in the literature [44,45], RDH techniques which operate in an encrypted medium [16,46] have become widely accepted due to their wide range of practical applications pertaining to the security of the cover medium.…”