Reversible Data Hiding Based on Multiple Adaptive Two-Dimensional Prediction-Error Histograms Modification

Abstract

Pairwise prediction-error expansion (pairwise PEE) is an efficient reversible data hiding technique based on two-dimensional (2D) histogram modification. However, without considering the image content, its embedding manner lacks adaptivity. In this paper, we propose the adaptive modification for multiple prediction-error histograms (PEHs). An iterative self-learning optimization algorithm is devised to adaptively generate the 2D mapping, based on the PEH and payload. A loss function is employed and the optimization can be solved linearly to reduce the time cost. To determine the appropriate expansion bins, we generate multiple 2D PEHs and establish multiple candidate mappings. The optimal combination of the adaptive 2D mappings is determined by using exhaustive searching. Our method is flexible as the 2D mapping generation has a broad dynamic range. The experimental results show that the proposed method can outperform the pairwise PEE and also provide a competitive performance compared with the other state-of-the-art 2D RDH methods.