基于混合DWT-DCT与帧分组的视频双域水印算法  

Video dual-domain watermarking algorithm based on hybrid DWT-DCT and frame grouping

作　　者：梁钰承 钮可[2,3] 李军[2,3] 孟逸飞[1,2,3] LIANG Yucheng;NIU Ke;LI Jun;MENG Yifei(Force Information Engineering University,Zhengzhou 450001,China;Key Laboratory of Information Security of the Chinese People’s Armed Police Force,Xi’an 710086,China;College of Cryptographic Engineering,Engineering University of the Chinese People’s Armed Police Force,Xi’an 710086,China)

机构地区：[1]信息工程大学,河南郑州450001 [2]武警部队信息安全重点实验室,陕西西安710086 [3]武警工程大学密码工程学院,陕西西安710086

出　　处：《网络与信息安全学报》2025年第1期106-128,共23页Chinese Journal of Network and Information Security

基　　金：国家自然科学基金(62202496,62272478)。

摘　　要：针对现有视频水印算法的鲁棒性与可逆性的兼容问题,提出了一种基于混合DWT-DCT与帧分组的视频双域水印算法。该算法基于H.264视频编码标准,提出一种视频双域水印框架,以视频帧的场景平滑度为标准对视频帧进行分组,在各分组中决策出符合标准的可嵌入帧作为隐秘帧用于水印嵌入,同时对隐秘帧进行混合DWT-DCT生成水印嵌入域。通过鲁棒嵌入方法将鲁棒水印嵌入Y通道低频嵌入域中;在Y通道高频嵌入域,通过直方图迁移技术嵌入可逆水印。最终在解码端实现水印的提取与信息完整性认证及原始视频的无损恢复。通过实验结果表明,所提算法具有良好的不可见性,各实验视频的嵌入帧的平均峰值信噪比与平均结构相似度的均值分别为49.009 2 dB与0.984 0,对不同强度的失真噪声攻击与帧攻击均具有强鲁棒,提取水印与原始水印的平均归一化互相关系数均在0.993 3以上,平均误码率均在0.058 2以下,同时计算复杂度低,实验视频单位帧编码的时间差均小于0.033 s,单位帧解码时间差均小于0.025 s,具有较高的应用价值。Aiming to address the compatibility of robustness and reversibility in existing video watermarking algo‐rithms,a video dual-domain watermarking algorithm based on hybrid DWT-DCT and frame grouping was pro‐posed.The algorithm,which was based on the H.264 video coding standard,introduced a video dual-domain water‐marking framework.Video frames were grouped according to their scene smoothness,and within each group,frames meeting the predefined criteria were selected as hidden frames for watermark embedding.The hidden frames were processed using a hybrid DWT-DCT approach to generate the watermark embedding domain.Robust watermarking was embedded into the low-frequency domain of the Y channel using a robust embedding method,while reversible watermarking was embedded into the high-frequency domain of the Y channel through histogram migration techniques.At the decoding end,watermark extraction and lossless recovery of the original video were achieved.Experimental results demonstrate that the proposed algorithm exhibits excellent invisibility,with the aver‐age peak signal-to-noise ratio and average structural similarity index of the embedded frames being 49.0092 dB and 0.9840,respectively.The algorithm is shown to be robust against distortion noise attacks and frame attacks of varying intensities.The average normalized correlation between the extracted watermark and the original water‐mark is above 0.9933,while the average bit error rate is below 0.0582.Additionally,the computational complexity is low,with the time difference for unit frame encoding and decoding being less than 0.033 s and 0.025 s,respec‐tively.These results indicates that the proposed algorithm has high application value.

关 键 词：视频双域水印框架 场景平滑度 混合DWT-DCT Y通道 H.264编码 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]