Reversible data hiding based on novel pairwise PVO and annular merging strategy

Abstract Reversible data hiding (RDH) methods aim to reduce the embedding distortion below the required embedding capacity. Concerning this problem, pixel value ordering (PVO)-based methods achieve remarkably good performance as they only modify the minimum and maximum of pixels in each block. In this paper, we further exploit the embedding potential of PVO by merging blocks that cannot be directly used in PVO with neighboring blocks, in order to generate a larger block. The mechanism of PVO is reconsidered and pixel distribution density (PDD) is defined as a metric to control both order and termination of merging strategy. Moreover, a novel two-dimensional prediction-error histogram is generated to enhance the embedding performance. Experimental results verify that the proposed method outperforms previous PVO-based schemes and state-of-the-art methods.