Abstract: In order to improve performance of color image Reversible Data Hiding（RDH）, a novel color image RDH method based on correlations of RGB channels is presented. At first, a RGB correlation based predictor is designed, including rough texture and precise texture prediction stages. Specifically, the first stage is to divide pixels into flat and texture regions by using one reference channel. In the second stage, the smooth pixel is classified into horizontal, vertical, smooth regions by using two reference channels so that the accurate prediction can be computed. The experimental results show the proposed predictor can accurately compute the prediction value and the data hiding performance is superior to the existing color image RDH methods.

Key words: color image, Reversible Data Hiding（RDH）, RGB channels, pixel prediction

摘要： 为了提高彩色图像可逆信息隐藏（RDH）的性能，提出了基于RGB相关性的彩色图像RDH方法。首先为了更加准确地预测像素值，设计了基于RGB相关性的像素预测算法。该算法包括粗糙纹理预测和精细纹理预测两个阶段。第一个阶段利用一个参考通道将像素分为平滑区域和纹理区域。在第二阶段，利用两个参考通道对平滑区域像素进行精细划分，将像素分为水平、垂直和平坦区域，从而更加准确地计算像素的预测值。其次，利用预测误差扩展技术嵌入秘密信息。实验结果表明提出的预测算法能够较准确地预测每一个通道的像素值，信息隐藏性能好于现有的彩色图像RDH方法。

关键词: 彩色图像, 可逆信息隐藏, RGB通道, 像素预测