基于宏块复杂度的自适应视频运动矢量隐写算法

发布时间：2018-12-11 04:25

【摘要】：信息隐藏是一种将秘密信息嵌入常见的载体中以实现信息秘密传递的技术.然而,隐写改动都会不可避免地造成各种失真.视频运动矢量隐写则会造成画面质量下降、比特率增加、概率分布改变等变化.这些由隐写造成的影响可能会被已有的或者潜在的隐写分析方法所检测.已有的运动矢量隐写分析特征大致可以分为两类:一类基于局部最优性特征,另一类基于时域和空域相关性特征.为了尽可能降低运动矢量的隐写失真,该文对视频编解码过程中运动矢量相关内容给出形式化描述方法.进一步地,理论分析了关于运动矢量的局部最优性以及相邻相关性两个因素.此外,文中研究了运动矢量隐写造成的最优概率下降的原因,指出了隐写纹理简单宏块的运动矢量不容易造成显著的局部最优性异常.该文讨论了运动矢量分量的分布规律,认为使得分量更靠近分布均值的修改,可以较好保持相关性.通过这些理论研究,提出基于宏块复杂度的视频运动矢量自适应隐写算法(Adaptive Macroblock Complexity,AMC),其可以分开处理运动矢量局部最优性和相邻相关性两个因素.该算法可以不使用编码实现秘密信息在最低代价路径上的嵌入和提取,而且每个运动矢量可以嵌入2比特秘密信息.算法包括两步:首先,压缩cover视频,不做任何嵌入,记录所有宏块压缩后的复杂度,同时依据分布统计设定一个复杂度阈值.然后,在第2次进行的视频压缩过程中,通过自适应阈值选择低复杂度宏块,将其运动矢量用于LSB(最低有效位)匹配嵌入.对于运动矢量来说,低复杂度宏块较容易保持局部最优,而LSB匹配嵌入可以保持相邻相关性.在实验部分,将不同隐写算法但相同嵌入率的stego样本在不可见性、码率增长、反检测能力等方面分别进行了对比.实验结果表明,cover的PSNR值最高,而AMC排在第二,与之相差不超过0.05dB.对比比特率增长,AMC算法比特率增长量只有其他算法的1%至50%左右.此外,该文使用两种隐写分析特征评估这些算法的安全性.检测的错误率表明AMC隐写算法和已有算法相比安全性更均衡.总体看来,AMC隐写算法能够较好抵抗隐写分析,特别是隐写后视频码率得到了控制,画面质量得到了很好保证.
[Abstract]:Information hiding is a technology which embeds secret information into common carriers to realize information secret transmission. However, steganography changes will inevitably lead to various distortions. The video motion vector steganography will result in the deterioration of picture quality, the increase of bit rate and the change of probability distribution. The effects of steganography may be detected by existing or potential steganography methods. The existing features of motion vector steganography can be divided into two categories: one is based on local optimality and the other is based on temporal and spatial correlation features. In order to minimize the steganography distortion of motion vectors, this paper presents a formal description method of motion vectors in video coding and decoding. Furthermore, the local optimality and adjacent correlation of motion vectors are analyzed theoretically. In addition, the reason for the decrease of optimal probability caused by motion vector steganography is studied, and it is pointed out that the motion vector of simple macroblock with steganography texture is not easy to cause significant local optimality anomalies. In this paper, the distribution of motion vector components is discussed, and it is considered that the correlation can be better maintained by modifying the components closer to the mean of distribution. Through these theoretical studies, a video motion vector adaptive steganography algorithm (Adaptive Macroblock Complexity,AMC) based on macroblock complexity is proposed, which can deal with local optimality of motion vector and adjacent correlation separately. The algorithm can embed and extract secret information on the least cost path without coding, and each motion vector can embed 2 bits of secret information. The algorithm consists of two steps: firstly, the cover video is compressed without any embedding, the complexity of all macroblocks is recorded, and a threshold of complexity is set according to the distribution statistics. Then, in the second video compression process, the low complexity macroblock is selected by adaptive threshold, and its motion vector is used for LSB (least significant bit) matching embedding. For motion vectors, low complexity macroblocks are easier to maintain local optimality, while LSB matching embedding can maintain adjacent correlations. In the experimental part, the stego samples with different steganography algorithms but the same embedding rate are compared in the aspects of invisibility, bit rate growth, anti-detection ability and so on. The experimental results show that the PSNR value of cover is the highest, while the AMC is the second, and the difference is not more than 0.05 dB. Compared with the increase of bit rate, the increase of bit rate of AMC algorithm is only about 1% to 50% of that of other algorithms. In addition, two steganographic features are used to evaluate the security of these algorithms. The error rate shows that the AMC steganography algorithm is more secure than the existing algorithms. In general, the AMC steganography algorithm can resist steganography better, especially the video bit rate is controlled after steganography, and the picture quality is well guaranteed.
【作者单位】：武汉大学空天信息安全与可信计算教育部重点实验室;武汉大学计算机学院;
【基金】：国家自然科学基金重点项目(U1536204);国家自然科学基金(61373169) 国家科技支撑计划项目(2014BAH41B00) 国家“八六三”高技术研究发展计划项目基金(2015AA016004)资助~~
【分类号】：TP391.41

【相似文献】