Reversible Data Hiding for Encrypted Image Based on Block Mean Difference Histogram Shifting

Reversible data hiding in encrypted image (RDHEI) is essential to protect data privacy. In this paper, we propose a novel RDHEI method based on block mean difference histogram shifting. The data owner divides the cover image into nonoverlapping 2×2 image blocks and computes the block mean difference. To ensure the accessibility of the block mean differences, the data owner encrypts the image block using Paillier while binding the mean differences to the encrypted image block using the self-binding property of Paillier. The data hider extracts the mean difference from the encrypted image blocks in the data embedding phase. The block mean difference histogram is constructed by calculating the differences based on the additive homomorphic properties of encrypted pixels within a block to enhance the density of the histogram distribution. Subsequently, the data is embedded into the encrypted image by histogram shifting to generate the encrypted marked image. Due to the homomorphic property of Paillier, the data receiver can simultaneously extract the data and restore the image from either the encrypted marked image or the plaintext marked image without loss, which achieves separability. Experimental results show that the proposed method performs great in terms of image quality and embedding rate, and its PSNR and embedding rate outperforms similar existing methods.