A Novel Reversible Data Embedding Scheme Using Dissimilar Pairing for Palette-Based Images

This paper presents a reversible data embedding scheme with prediction and dissimilar techniques for palette-based images. Two embedding rules are imported in the proposed method, and no extra bit map is required to indicate which rules are applied in which pixels to embed secrets. As the experimental results show, when the hiding capacity is 255 Kb, the visual quality of the decoded image for our proposed method is, on average, better than Fridrich’s method, Tzeng et al.’s method, or Chan and Chang’s method, which are 4 dB, 3 dB and 2 dB, respectively. With restoration ability, the visual quality of our proposed scheme outperforms these methods proposed previously.