Location-Based PVO and Adaptive Pairwise Modification for Efficient Reversible Data Hiding.
Tong Zhang,Xiaolong Li,Wenfa Qi,Zongming Guo
DOI: https://doi.org/10.1109/tifs.2019.2963766
IF: 7.231
2020-01-01
IEEE Transactions on Information Forensics and Security
Abstract:Pixel-value-ordering (PVO) is an efficient technique of reversible data hiding (RDH). By PVO, the maximum and minimum in each cover image block are first predicted and then modified to embed data. Actually, many PVO-based methods are essentially based on high-dimensional histogram modification. For these methods, a two-dimensional (2D) prediction-error histogram (PEH) is first generated and then modified based on a 2D mapping. However, these methods have two drawbacks. On one hand, the generated 2D PEH is irregular so that it is difficult to design suitable histogram modification strategy. On the other hand, the employed 2D mapping is empirically designed, and thus the embedding performance is far from optimal. Based on these considerations, a new PVO-based RDH scheme is proposed in this paper. By considering both pixel value orders and pixel locations, a new predictor is proposed so that the generated 2D PEH is regular in shape and suitable for reversible embedding. Moreover, instead of manually designing 2D mappings, to optimize the embedding performance, a self-learning mechanism is proposed to adaptively select the 2D mapping according to the image content. With the new predictor and the self-learning mechanism for 2D mapping selection, the proposed method works well with a good marked image quality, e.g., the PSNR of the image Lena is as high as 61.53 dB for an embedding capacity of 10 000 bits. Besides, compared with some state-of-the-art RDH methods, the superiority of the proposed method is experimentally verified.