摘要：
随着视频编辑软件功能的增强,视频内容的篡改和伪造变得越来越容易,在很多领域可能导致严重后果.在视频依次经历解码、常规信号处理及修改图像组GOP(Group of Pictures)与量化参数QP(Quantization Parameters)的二次压缩条件下,如何有效判定视频内容的真实性是个亟待解决的问题.因此,该文提出了一种基于调制奇异值贡献率的视频内容认证水印算法:(1)通过在视频亮度分量子块的2级DWT(Discrete Wavelet Transform)域抽取特征与嵌入水印,能从整体上提升重构的特征与恢复的水印的一致性;(2)选择在2级DWT的低频域,基于量化调制第一奇异值贡献率的新颖算法来嵌入水印,实现了水印的透明性和抗二次视频压缩的鲁棒性;(3)基于能量关系所表达的视频内容特征,能在二次压缩条件下保持高度的稳定性;(4)采用水印交叉嵌入法与邻域判定法,提升了算法的安全性及判定的准确性.理论分析表明,提出的算法能抵制伪造、拼贴等攻击,在无恶意篡改的二次压缩条件下,水印的鲁棒性为0.96.实验结果表明,水印的嵌入对视觉质量的影响平均下降1.3％,对码率的影响平均增加8.41％,提出的算法在H.264/AVC二次压缩及期间的常规信号处理条件下能有效区分视频保持内容的处理与恶意篡改,并定位篡改位置.%With the function enhancement of video editing software,video contents tampering and forging become more and more easier,which makes the user difficult to distinguish the authenticity of video content.So it may lead to serious consequences when it comes to forensic evidence,medical expertise,military intelligence,national security,and so on.Up to now,most of the existing researches focus on the identification of video content under the first compression or recompression.But after videos are published in compressed form,it often undergo decoding,conventional signal processing,and encoded again by modifying GOP(Group of Pictures) and QP (Quantization Parameters).Under the secondary compression condition above,how to effectively determine the authenticity of video content is an urgent problem.Therefore,we propose a watermarking algorithm,based on quantitatively modulating the contribution rate of the singular value,for video content authentication.(1) By extracting features and embedding watermark in the 2nd level DWT (Wavelet Transform Discrete) domain for each sub-block from the video luminance component,the consistency between the reconstructed features and extracted watermark can be promoted in the whole;(2) The low frequency domain in the 2nd level DWT is selected to embed the watermark by using a novel algorithm based on quantitatively modulating the contribution rate of the first singular,which achieves watermark transparency and the robustness tolerating the second video compression.The core content of authentication algorithm based on digital watermarking is the stability of hidden information which is embodied by watermark algorithm.Because the stability of the proposed watermark comes from all the singular values,rather than part of the singular value,algorithm of this paper have obvious advantages in robustness;(3) The video content features based on energy relations can maintain a higher stability after the secondary compression;(4) The security and accuracy of the algorithm are improved respectively by crossing watermark and neighborhood decision.In addition,in this paper,the security of the proposed algorithm is analyzed theoretically,including cryptanalysis,collage attack,quantization attack and targeted forgery attack.And the theoretical analysis shows that the proposed algorithm can resist forgery attack,collage attack and some other attacks,and the robustness of the watermark is 0.96 under the condition of the secondary compression without malicious tampering.Finally,the experiment is carried out aiming at the transparency of watermark,the impact on bit rate and the robustness of attack.And the experimental results show that the influence of watermark embedding on visual quality is decreased by 1.3％,and that on bit rate is increased by 8.41％;that the proposed watermarking scheme can effectively distinguish malicious manipulations from the content-preserving operator after secondary compression for H.264/AVC video,and locate the position tampered.Also,through the performance comparison of related schemes,it further demonstrates the advantage of this algorithm for video content authentication under the condition of secondary compression.
