Research On Robust Reversible Watermarking Technology Against JPEG Compression

Digital watermarking technology mainly embeds private information into carrier information for transmission,and can ensure that it does not affect the normal use of carrier information.As an important branch of information security,this technology is one of the important ways to realize copyright protection and covert communication.Reversible watermarking and robust watermarking are two mature research points in digital watermarking technology.Reversible watermarking guarantees the recovery of carrier image and the extraction of watermarking information after lossless channel transmission.Robust watermarking emphasizes the robustness of watermarking algorithm.Robust reversible digital watermarking(RRW)has the advantages of both reversible digital watermarking and robust digital watermarking.In lossless channel,the watermarking algorithm is reversible;in lossy channel,the watermarking algorithm is robust.But nowadays,most of the researchs on robust reversible watermarking focus on lossless images,and there is not much to do with the more widely used lossy images(such as JPEG images,Joint Photographic Experts Group Joint Image Experts Group).After studying the existing robust reversible watermarking algorithms and the whole process of JPEG compression,this thesis proposes a robust reversible watermarking algorithm for JPEG format pictures against JPEG compression.In the whole process of JPEG compression,only quantization will result in the loss of image energy.Therefore,the DCT coefficients(DCT Discrete Cosine Transform)after quantization in JPEG compression process are used as the embedding carrier of watermarking when the algorithm actually operates.Based on the existing algorithms,the multi-stage embedding technique,which embeds watermarking information and distortion auxiliary information in two stages respectively,is proposed in this thesis.To solve the problems of existing algorithm,including the poor performance of the algorithm caused by the dependence of the two embedding domains and the inadaptability of the algorithm to JPEG images,our algorithm has been greatly improved and innovated.By studying the attributes of DCT coefficients at different locations,the algorithm divides the coefficients into three categories: low frequency,medium frequency and high frequency.The specific watermarking will be embedded in the medium region coefficients which have little impact on the carrier image and show robustness to various attacks.In order to guarantee the independence of the two embedding domains,the medium frequency coefficients are divided into the mid-low frequency regions suitable for watermarking embedding and the mid-high frequency regions suitable for distortion aided information embedding.The first stage of the algorithm is to embed watermark information by modifying the mid-low frequency coefficients.The second stage of the algorithm is embedding the distortion auxiliary information through shifting mid-high frequency coefficient histogram.In order to increase the robustness of the algorithm and the visibility of the watermarking image,a coefficient selection method based on block complexity is proposed.Specifically,the watermarking information will be embedded in the mid-low frequency coefficients of complex blocks,and the recovery information will be embedded in the the mid-high frequency coefficients of smooth blocks.In this thesis,the bit error rate of watermarking after attack is used as a performance index,and our algorithm is compared with the contrast algorithm under the same experimental conditions with different attack types and intensity.Experiment results show our method performs better than some state-of-the-art RRW methods.Meanwhile,in order to ensure the requirement of image size for JPEG compression,the size of image before and after watermarking embedding is compared.Experiments show that our method has little influence on the size of JPEG image.