PVO-based reversible data hiding using adaptive multiple histogram generation and modification

A recently proposed pixel-value-ordering (PVO)-based reversible data hiding (RDH) method takes a sharp histogram derived from two skewed histograms, which realizes good performance through reducing the number of shifted pixels. However, in this recent work, since pixels with different local complexity are processed with the same modification manner based on a single histogram without considering their different local properties, the embedding performance is far from optimal. In this paper, to better exploit the pixel complexity and enhance the reversible embedding performance, a novel PVO-based RDH method using adaptive multiple histogram generation and modification is proposed. First, pixels with different local complexity are divided into several classes utilizing multiple thresholds. Then, a PVO-based predictor is used for prediction and multiple prediction-error histograms corresponding to the different pixel classes are obtained. Next, the generated histograms are modified to embed data according to their statistical characteristics. Here, based on the established capacity-distortion model, the histogram generation and modification are processed in an adaptive way to optimize the embedding performance. Moreover, by extending the proposed method to multiple two-dimensional prediction-error histograms, the embedding performance can be further improved. Experimental results verify that the proposed method outperforms certain state-of-the-art techniques with good marked-image quality.