Abstract: In order to solve the shortage of employing hierarchical relationships of original images in reversible data hiding based on difference histogram shifting, this paper develops a novel reversible data hiding algorithm based on hierarchical structure and difference histogram shifting RDH-HSDHS. Firstly RDH-HSDHS uses the pixel differences of original image’s blocks to form difference histogram, and then utilizes the high correlations between neighbor pixels to embed data. In order to employ the reference pixels in blocks further, RDH-HSDHS rearranges these reference pixels to form the next embedding layer. This multiple-layer embedding procedure will be stopped until the current layer’s embedding capacity is less than the side information for decoding or the output image’s quality is lower than a given threshold value. Experimental simulation results demonstrate that the proposed algorithm makes use of the overall and local characteristic of original image for obtaining the larger number of secret bits for embedding, with good quality of the output images. When the marked image’s quality is larger than 30 dB, the embedding capacity can reach more than 0.75 bit/pixel. In addition, the proposed algorithm outperforms conventional methods based on difference histogram shifting. The presented algorithm is compatible to carry out secret information hiding in digital image with high quality requirement.

Key words: reversible data hiding, hierarchical structure, difference histogram, histogram shifting

摘要： 针对现有基于差值直方图方法利用原始图像结构关系上的不足，提出了一种基于层次结构和差值直方图平移的无损数据隐藏方法RDH-HSDHS。RDH-HSDHS利用原始图像数据块中像素的差值形成直方图，充分利用图像中相邻像素间的相关性嵌入数据，为了进一步利用数据块中的参考像素进行数据嵌入，将参考像素组成新的图像进行下一层水印嵌入，直到当前层的嵌入容量小于解码所需的附加信息的长度或隐秘图像质量小于给定阈值。实验仿真结果表明，RDH-HSDHS能较好利用原始图像的全局和局部特性，在嵌入容量和隐秘图像质量之间达到较好的折中，在隐秘图像质量超过30 dB的同时，嵌入容量大于0.75 bit/pixel。另外，与相似方法的性能比较证明了提出方法的优势。该方法能有效应用于高质量需求的图像载体中进行信息隐藏。

关键词: 无损数据隐藏, 层次结构, 差值直方图, 直方图平移