摘要： 传统基于预测误差直方图平移的可逆信息隐藏算法大多通过固定顺序来扫描原始图像，从而进行数据嵌入，这种方式没有考虑图像本身的纹理信息，导致无效移位像素点较多，伪装图像视觉质量较差。为解决该问题，提出一种基于中值预测的四轮嵌入可逆信息隐藏算法，以在提高嵌入容量的同时降低伪装图像的失真率。利用相邻像素之间具有较强相关性的特点，在较小的误差值处聚集大量像素点，以得到更陡峭的预测误差直方图并提高嵌入容量。对每个像素点定义复杂度，根据复杂度的高低对预测误差进行排序，优先在图像平滑区域嵌入数据，从而有效减少无效移位像素点个数，降低伪装图像的失真率。实验结果表明，该算法的最大嵌入率可以达到0.3 bpp，在0.1 bpp的嵌入率下峰值信噪比高达55.15 dB，与非对称直方图算法、误差直方图移位算法等相比，其具有较高的嵌入容量和较小的视觉失真率。

关键词: 可逆信息隐藏, 预测误差, 直方图平移, 嵌入容量, 中值预测

Abstract: Traditional reversible information hiding algorithms based on prediction-error histogram translation primarily use a fixed scanning sequence to scan the original image for data embedding.This method does not consider the texture information of the image itself, which results in a greater number of invalid shifted pixels and poor visual quality of the camouflage image.To solve this problem, a four-round embedding reversible information hiding algorithm based on median prediction is proposed to improve the embedding capacity and reduce the distortion rate of the camouflage image.This study takes advantage of the strong correlation between adjacent pixels to gather many pixels at the smaller error value.This results in a steeper prediction error histogram and improves the embedding capacity.The complexity is defined for each pixel, the prediction error is sorted based on the complexity, and the data are preferentially embedded in the image smoothing area to reduce the number of invalid shifted pixels and reduce the distortion rate of the camouflage image.Experimental results show that the maximum embedding rate of the algorithm can reach 0.3 bpp, and the Peak Signal-to-Noise Ratio(PSNR) is as high as 55.15 dB at an embedding rate of 0.1 bpp.Compared with the asymmetric histogram algorithm and error histogram shift algorithm, the proposed algorithm has a higher embedding capacity and smaller visual distortion rate.

Key words: reversible information hiding, prediction error, histogram translation, embedding capacity, median prediction

中图分类号: 

• TP309