基于最近邻像素预测的加密域可逆信息隐藏

针对现有的可逆信息隐藏算法在图像加密域应用场景下嵌入容量不高的问题，提出了一种基于最近邻像素预测的加密域可逆信息隐藏算法。充分利用自然图像的空间相关性，设计最近邻像素预测方法，计算预测误差，并进一步探究预测误差块分类规律，设计块内移位以及块重排方式，保留了块内相邻像素相关性。然后，通过动态标记块内误差区间以最大限度地空出冗余空间，达到提高嵌入容量的目的。选用5幅标准灰度图像及数据集BOSSbase、BOWS-2和UCID中的万余张灰度图像进行对比实验。结果表明：本文算法在安全性、可逆性和可分离性都满足的前提下，嵌入容量相比同类算法平均提升0.35～0.9 bpp(比特数)。

Reversible Data Hiding in Encrypted Images Based on Nearest Neighbor Pixel Prediction

Aiming at the problem that the embedding capacity of existing reversible data hiding scheme in encrypted image was not high, a reversible data hiding in encrypted images based on nearest neighbor pixel prediction was proposed. By making full use of the spatial correlation of natural images, the nearest neighbor pixel prediction method was designed, and the prediction error was calculated. Furthermore, the classification rule of prediction error blocks was further explored, and the intra-block shift and block rearrangement were designed to preserve the correlation of adjacent pixels within blocks. Then, the error range within the block was dynamically labeled to maximize the redundancy space, so as to improve the embedding capacity. In this paper, comparative experiments are carried out on five standard images and more than ten thousand gray images from public dataset (BOSSbase, BOWS-2 and UCID). The experimental results show that the embedding capacity of the proposed scheme is improved by 0.35-0.9 bpp on average compared with similar algorithms under the premise that the security, reversibility and separability are satisfied.