High-Fidelity Reversible Data Hiding in JPEG Images Based on Two-Dimensional Histogram

Recently, a series of excellent reversible data hiding (RDH) algorithms based on two-dimensional histogram modification have been proposed. However, the coefficient-pairs of these algorithms are obtained via fixed combination method, which limits the further improvement of algorithm performance. Based on this consideration, a new combination method with pairwise coefficients is proposed in this paper. Firstly, the coefficients are divided into 32 positions according to Zigzag scanning sequence. Then, we pair the coefficients in the same position, and combination mode of the coefficients is determined by the number of expandable pairs. Unlike the fixed combination method, the coefficients of proposed method are adaptively combined into pairs according to image features. Moreover, we propose a coefficient-pairs selection strategy. The coefficient-pairs which have little effect on image quality are selected for payload embedding, thereby distortion of the mark image is limited. The obtained algorithm has better peak signal to noise ratio (PSNR) and embedding capacity over some state-of-the-art algorithms, showing the superiority of our method.