摘要： 为提高加密图像可逆数据隐藏的嵌入容量与安全性，提出一种改进的加密域可逆数据隐藏算法。以分块加密方式保留图像块内相邻像素间的相关性，利用多元线性回归模型对图像块内特定像素进行预测，同时以位替换方式将预测误差码元和秘密信息同时嵌入目标像素中，通过差分对称编码提升编码效率并扩展嵌入空间，使得接收者在仅拥有加密密钥和嵌入密钥的情况下可无失真地恢复原始图像和提取秘密信息。实验结果表明，该算法在BOWS-2和UCID数据集上的平均嵌入率分别达到1.717和1.310 bit/pixel，在提高嵌入容量和安全性的同时实现了信息提取操作和图像恢复操作的完全分离。

关键词: 多元线性回归, 加密域, 可逆数据隐藏, 差分对称编码, 嵌入容量

Abstract: To improve the embedding capacity and security of the Reversible Data Hiding in Encrypted Images(RDH-EI),this paper proposes an improved Reversible Data Hiding(RDH) algorithm in encrypted domain.In the algorithm,block-level encryption is used to keep the correlation between adjacent pixels in the image block.Then,the specific pixels in the image block are predicted by using the multivariable linear regression model.At the same time,the encoded prediction error and secret information are embedded into the target pixels in the way of bit replacement,and the method of differential symmetric coding is used to improve the coding efficiency and the embedding capacity.Thus the receiver can restore the original image without distortion in the case of having only the encryption key,or extract the secret information in the case of having only the embedded key.Experimental results show that the proposed algorithm effectively improves the average embedding rate,which reaches 1.717 bit/pixel on BOWS-2 and 1.310 bit/pixel on UCID,demonstrating its high embedding capacity and security.Also,it enables information extract operations to be fully separable from the image restoration operations.

Key words: multiple linear regression, encrypted domain, Reversible Data Hiding(RDH), differential symmetric coding, embedding capacity

中图分类号: 

• TP309