【作者】 肖尚勤；

【导师】 卢正鼎； 凌贺飞；

【作者基本信息】 华中科技大学 ， 计算机应用技术， 2008， 博士

【摘要】 鲁棒数字水印技术在多媒体内容保护和相关领域中具有很好的应用价值,如版权保护、盗版追踪、拷贝控制和广播监控等等。但目前鲁棒数字水印技术还未得到充分的研究,有很多技术方面的难题亟待解决,这给研究工作带来了挑战和机遇。对数字水印的深入研究将侧重于完善相关理论,提高水印算法的鲁棒性,增强水印技术的安全性。数字水印技术的两大重要性质是不可感知性和鲁棒性,一个好的鲁棒水印算法应该在不可感知的前提下能够抵抗各种攻击,但不可感知性和鲁棒性是相互矛盾的,鲁棒水印研究的关键之一在于如何选择合适的视觉感知模型,达到水印鲁棒性和不可感知性的平衡与统一。大量的水印算法主要利用扩频、信道编码和视觉感知模型来提高水印的鲁棒性,这对一般信号处理具有较好的鲁棒性,但难以抵抗由旋转、缩放等几何变换造成的去同步攻击,抗几何攻击的水印算法是需要解决的难点问题。此外,在视频水印中还要考虑时间轴上的攻击,如帧平均、帧删除、帧重组等,因此针对不同的应用需求设计鲁棒水印是水印研究的重点。与Contourlet变换相结合的鲁棒水印算法(AWCD),分析Contourlet变换特性,在此基础上建立相应视觉掩蔽模型,综合频率、亮度和纹理参数确定水印嵌入强度,自适应地将水印信息嵌入到图像Contourlet分解后的高频子带,保证了水印的不可见性,分别对水印和嵌入区域高频子带系数置乱以提高水印的安全性和鲁棒性,水印信息在Contourlet逆变换中影响到低频子带,使水印分布到整个载体图像的高频和低频区域,因此AWCD算法具有较强的鲁棒性,水印的提取通过相邻系数对比的盲检测算法,能够不借助原始图像提取水印。针对抗几何攻击的数字水印研究,提出基于平衡多小波的视频水印算法(VW-BMW),将载体视频转换到对数极坐标变换系LPM,使旋转和缩放转换为该坐标系下的平移,构造基于平移不变的平衡多小波,得到具有旋转、缩放不变性的能量特征值,通过修改该值嵌入水印,使算法对信号处理攻击和几何攻击都具有鲁棒性。所构造的平衡多小波不需要预滤波,计算耗费较少,能够满足视频水印所需要的实时性。水印嵌入时根据嵌入区域的能量特征值自适应控制水印的嵌入强度,保证水印嵌入达到最大强度的同时不会造成视觉失真。在小波域数字水印算法研究的基础上,结合Shamir门限密钥分享算法提出视频水印算法(SSVW),通过拉格朗日插值多项式将水印图像分为多个水印影子,循环嵌入到视频序列中。在嵌入水印时,对水印影子的不断更新,使算法能够有效抵抗共谋攻击,在提取水印时,根据密钥分享理论利用部分水印影子信息就能够恢复原始水印,为了防止欺骗攻击,采用单向散列hash函数验证水印影子的有效性。SSVW算法不但对常见信号处理和攻击有较好的鲁棒性,而且对于视频中的帧删除、帧平均攻击的情况,也能很好的检测出水印。更多还原

【Abstract】 Robust digital watermarking is widely used in multimedia content protection and correlation field, such as copyright protection, traitor tracing, copy protection and broadcast monitoring. But the technology of video watermarking hasn’t been developed enough. There are still lots of problems to be solved which bring chances and challenges for study. The study of digital watermarking will emphasize the robustness and security of watermarking system.The important characters of watermarking are robustness and imperceptibility. But they are contradictive indeed. The key of this problem is how to make an appropriate Human Visual System to achieve the balance of robustness and imperceptibility. At present, the most of watermarking algorithms increase robustness of the watermarking by using the spread spectrum, the channel coding and the performance of HVS. These algorithms are robust to the commonly signal processing, but they are difficult to resist the geometric attacks such as rotation and scale. Therefore, how to solve the problem of geometric attacks in embedding and detecting is a still difficult research. Moreover, video sequence may be faced with some temporal attacks, such as frame average, frame removal, frame reconstruct and so on. So the research of robust watermarking for different application is valuable.The Contourlet transform is adopted by virtual of its advantages of a flexible multiresolution, local and directional expansion because of using contour segments. The robust watermarking algorithm in Contourlet domain (AWCD) was proposed. According to the frequency, luminance and texture characteristics of HVS, the watermark composed of a pseudorandom sequence is adaptively embedded into the high-pass subbands. The detection of watermarking is according to the neighboring coefficients, and the algorithm is blind. Experimental results show that the algorithm has good robustness against attacks and imperceptibility.According to that the balanced multiwavelets are simultaneously orthogonal, symmetric and compactly supported the watermarking algorithm (VW-BMW) is proposed based on BMW. The algorithm combines the LPM to get an invariant energy for rotation, and scaling. We embed the watermark into the energy features to achieve the robustness resist geometric distortions. The balanced multiwavelets need no prefiltering procedure and have show better prospects in application especially in computing complexity. So it can satisfy the real-time and robustness of video watermarking simultaneously.A robust and blind video watermarking (SSVW) based on Shamir’s secret sharing scheme for multiusers is proposed. It divides the field of host video signal into multisegments and then splits the original watermark into corresponding number shares with secret sharing algorithm. Each share of watermark is embedded into the corresponding segment of host video signal. The computing complexity of algorithm is reduced by grouping the frame sequence. In the embedding processing, the updating of watermark shadows is periodic, so the scheme proposed can resist collusion attack. In the extract processing, the hashing function is used before watermark reconstruction to resist the cheating attack. Experiment results show that the proposed watermarking technique is robust against many signal processing especially the frame average and frame lose.更多还原