Reversible data hiding using Delaunay triangulation and selective embedment

In this paper, we propose a reversible data hiding method based on prediction and histogram-shifting (PHS) using Delaunay triangulation and selective embedment. Because the statistical properties of the prediction errors along the embedding direction are significantly altered due to the shifting operation, the existing PHS-based methods are vulnerable to some steganalyzers. The proposed method exploits a set of key-selected referential pixels to construct a 3D Delaunay mesh to obtain the prediction errors. The presence of the altered statistical properties along the embedding direction is unrevealed because the same set of prediction errors cannot be reconstructed. A selective embedment mechanism is used to control the embedding regions in the cover image for evading the detection of the steganalyzers based on low-amplitude stego signal. The experimental results reveal that the proposed method not only provides better payload and image quality than the existing PHS-based methods, but is also robust to the detection of modern steganalyzers such as histogram analysis and SPAM.