基于运动矢量的视频隐写算法

视频信息隐藏是一种将秘密信息嵌入到视频中实现隐蔽传输的技术.基于运动矢量(motion vector,MV)的视频隐写造成的运动矢量特性异常容易被隐写分析所检测,隐蔽通信的安全性降低.为了保持运动矢量相邻相关性和局部最优性,提出了一种基于运动矢量的视频隐写算法能够降低对运动矢量特性的破坏.首先,实验验证已有基于运动矢量的视频隐写算法对运动矢量不同分量的影响存在差异.然后,引入归一化互信息(normalized mutual information,NMI)衡量运动分量的相邻相关性,根据载体选择策略选择相邻相关性小的分量作为嵌入载体.最后,定义宏块纹理复杂度作为载体修改的失真代价,利用校验格编码(syndrome-trellis code,STC)对载体进行信息嵌入.实验结果表明,隐写视频的峰值信噪比(peak signal to noise ratio,PSNR)和结构相似性(structural similarity,SSIM)降低不超过0.4％,一般低于0.25％;比特率增加0.29％ ～2.35％;算法被相关异常特征检测的错误率较高,具有更均衡的安全性.可见,算法各项性能指标表现良好,具有较好的抗隐写分析能力,隐写视频不可感知性较好.

Video Steganography Algorithm Based on Motion Vector

Video steganography is the art of covert communication which sends secret message under the video. However, abnormal motion vector (MV) characteristics introduced by MV-based steganography may be detected by anti-steganalysis and the security of covert communication will be diminished. An MV-based video steganography algorithm was proposed in this paper to reduce the damage for the characteristics of MVs includes the adjacent correlation and local optimality. First, the MV-based video steganography algorithms have a different impact on the abnormality of different MV components was verified experimentally. Next, the normalized mutual information (NMI) was introduced to evaluate the adjacent correlation of the MV components and the small one was selected as the steganographic carrier according to the carrier selection strategy. Finally, the complexity of the macroblock was defined as the distortion cost of the steganographic carrier modification and STC (Syndrome-Trellis Code) was employed to embed the secret message. Experimental results show that the PSNR and SSIM of stego video diminished no more than 0.4% at worst, generally less than 0.25%, the bit rate increased range 0.29 to 2.35%, the detection error rate of the proposed algorithm by anti-steganalysis shows that our algorithm has more balanced security than others. Overall, the performance indicators of the proposed algorithm perform well and have a good ability to resist anti-steganalysis, stego video has high imperceptibility.