A Novel Reversible Data Hiding Scheme Based on Reference Pixels and Adaptive Block Selection

In this paper, a novel reversible data hiding (RDH) scheme based on reference pixels and adaptive block selection is proposed. We first divide the cover image into non-overlapping blocks. Then for each block, we assign the pixels with median value as reference pixels and the number of reference pixels as a smooth level (SL). In this way, residual histograms for each SL can be generated. For the blocks with SL not less than a given threshold, we embed data into them using histogram shifting. Here, we embed data from the blocks with the highest SL (the most smooth ones) to lower ones instead of sequential embedding. For the other blocks, each of them is further divided into two new ones containing the pixels larger or less than its median. These new blocks will be embedded with the rest secret data. By the proposed scheme, it can adaptively embed data into the smooth blocks and less modification is made to noisy ones. The experimental results demonstrate our superiority over some state-of-the-art RDH works.