Pixel-value-ordering (PVO) is an effective and promising method of reversible data hiding (RDH) and has received much attention in recent years. To improve performance, a pixel-based PVO (PPVO) method was recently introduced to predict the pixels to be embedded in a pixel-wise manner instead of the block-wise manner used by PVO. However, for PPVO, the surrounding neighbors of the predicted pixels are underutilized; moreover, its embedding does not adapt to the local complexity of the image to be embedded. To overcome the shortcomings of PPVO, this paper proposes a novel PVO method based on hybrid prediction for RDH. First, the surrounding neighbors of the pixel to be predicted are fully utilized by a hybrid prediction method, which combines rhombus prediction and pixel-wise prediction. Second, a modified embedding scheme based on multiple histograms is presented for adaptive embedding. Experimental results show the superior performance of the proposed method by comparing it with state-of-the-art RDH methods.

像素值排序（PVO）是一种有效且有前途的可逆数据隐藏（RDH）方法，近年来受到了广泛关注。为了提高性能，最近引入了基于像素的PVO（PPVO）方法，以像素方式而不是PVO的逐块方式预测要嵌入的像素。但是，对于PPVO，未充分利用预测像素的周围邻居。此外，其嵌入不适应要嵌入图像的局部复杂性。为了克服PPVO的不足，提出了一种基于混合预测的RDH新的PVO方法。首先，将菱形预测和逐像素预测相结合的混合预测方法充分利用了要预测像素的周围邻居。第二，提出了一种基于多个直方图的改进嵌入方案，用于自适应嵌入。通过与最新的RDH方法进行比较，实验结果表明了该方法的优越性能。