Adaptive multi-histogram reversible data hiding with contrast enhancement

Abstract

Unlike existing reversible data hiding with contrast enhancement (RDHCE) methods, which excessively improve the image contrast for achieving the required capacity, the proposed method improves the image contrast appropriately while providing satisfactory embedding capacity. To this end, an adaptive multi-histogram RDHCE method is proposed in this study to improve the local and global contrast by considering the local properties of the histograms. On the one hand, fuzzy C-means clustering combining multiple features that are deliberately designed for contrast enhancement is employed to generate seven sharply-distributed prediction error histograms (PEHs). Subsequently, the genetic algorithm is utilized to adaptively select the optimal pairs achieving the best embedding performance for each PEH according to the local characteristics of PEH distribution, resulting in improving the local contrast adaptively and embedding significant amount of data. Additionally, two-sided histogram shifting (HS) is utilized to improve the global contrast appropriately while embedding reasonable amount of data. The experimental results demonstrate that the proposed method achieves better local and global contrast while providing a high embedding capacity compared with other existing RDHCE methods.