基于分布式信源编码和边信息优化机制的密文域图像可逆信息隐藏研究

密文域可逆信息隐藏在密文数据中嵌入信息并能够无损恢复原始明文载体，可实现保护内容所有者隐私条件下的数据标注、管理和追踪，也是学术界研究的热点。本项目针对数字图像开展密文域可逆信息隐藏的理论模型和方法研究。为获得给定解密图像失真条件下的密文图像嵌入量的理论极限，本项目首先借助多用户信息论将密文图像可逆嵌入等价为接收端具有边信息条件的密文图像无损压缩的一般性问题，建立基于分布式信源编码和边信息优化机制的密文域可逆隐藏的理论模型。并在此基础上实现方法创新：通过基于图像内容显著性特征的自适应参考比特分配方案增加边信息容量，提出采用渐进式密文图像无损压缩的可逆信息隐藏新方法，以提高率失真性能；构建根据接收端已解码数据的信源模型统计特性估计和嵌入率弱反馈机制，动态调整信息隐藏者的密文图像压缩率和信息嵌入率，有效保证信息提取和图像恢复的可靠性。项目成果对多媒体信息安全和隐私保护具有重要意义。

Reversible data hiding in encrypted domain can embed additional information into cipher data sent from the content owner on the data-hider side and recover the original plaintext data on the receiver side, which can effectively realize data labeling, management and tracking without revealing the plaintext content and sacrificing the privacy of content owner. The topic has attracted the significant interests among the academia recently. This project will focus on the research of theory and methodology for reversible data hiding in encrypted images (RDHEI). First, in order to obtain the theoretical limit of hiding capacity in encrypted image under the conditional of the given distortion for decrypted image, we will define the RDHEI as the general problem of lossless compression for encrypted image with side information, and establish the theoretical model for RDHEI based on the distributed source coding and side-information optimization mechanism. According to this theoretical model, we will carry on the technical innovations for RDHEI. Through adaptively assigning the reference-bits of salient features and increasing the capacity of side information, we will propose novel RDHEI methods by progressive lossless compression to improve the rate-distortion performance. Also, we will study the mechanism of source statistics estimation based on decoded data and weak feedback of embedding rate, and dynamically optimize the rates of compression and embedding on the data-hider side to guarantee the accuracy of data extraction and image recovery. The research results of this project will have the important significance for multimedia security and privacy protection.