Research On A High-capacity Robust Data Hiding In Encrypted Images Based On Compressive Sensing

With the rapid development of multimedia signal processing and network technology,the protection of visual privacy has become an important issue in multimedia security.Therefore,data hiding in encrypted images(DHEI)has become a hotspot that has attracted much attention.However,the existing DHEI schemes mostly focus on the lossless recovery of the cover image,that is,reversible data hiding in encrypted images,and seldom pays attention to the embedding capacity and robustness of the scheme.A superior DHEI scheme needs to consider the balance among security,robustness,embedding capacity and recovered image quality.Compressive sensing(CS)is an emerging signal sampling and processing technology that breaks the limitations of Nyquist Sampling Theorem.It can sample data at an extremely low sampling rate and reconstruct the original signal with extremely high accuracy.CS realizes data sampling,compression,and encryption at the same time.These features are consistent with the purpose of data hiding in encrypted images,which brings it huge potential in the field of data hiding in encrypted images.This thesis makes a preliminary exploration of the coordination between CS and DHEI.To some extent,this thesis utilizes CS to achieve the balance of various indicators of DHEI.That is,it improves the embedding capacity and robustness of DHEI scheme while ensuring the quality of recovered image.The main content of this thesis includes:(1)A high-capacity DHEI scheme based on CS progressive reconstruction and predictive coding is proposed.Based on the democratic characteristics of CS measurements,the scheme innovatively proposes a progressive reconstruction method for CS and uses it to realize Most Significant Bits(MSBs)prediction of image data,thereby creating a large amount of usable space for data hiding.Experimental results show that compared with other existing related works,this scheme has a higher embedding capacity,and can ensure stable recovery of cover images with high visual quality under various embedding capacities.(2)Aiming at the flaws in the robustness of the first scheme,a robustness improvement scheme based on scrambling and Kronecker CS is proposed.The scheme makes full use of the encryption and compression characteristics of CS measurements to complete the encryption and pre-processing of the image at the same time,and utilizes MSBs replacement,image global scrambling and adjusted uniform quantization to improve the robustness of the algorithm.On the other hand,two-dimensional Kronecker CS reduces the computational complexity of the encoding end and improves the encoding efficiency.The simulation experiment results show that the scheme maintains the advantages of high-capacity embedding and stable recovery.In addition,it can extract data with a certain probability and restore the original cover image under noise attack and image cropping attack.