【作者】 王晓旭；

【导师】 费耀平；

【作者基本信息】 中南大学 ， 计算机应用技术， 2009， 硕士

【摘要】 随着网络技术与压缩技术的发展,人们可以方便快捷地复制和传输各种数字视频。如何对大量的数字视频进行有效的版权保护成为一个迫切需要解决的问题,视频数字水印技术为这一问题提供了一个有效的解决手段。本文对视频数字水印技术进行了研究,主要在以下几个方面取得了一些研究成果:提出了一种基于DCT-SVD域的原始视频数字水印算法。它利用视频场景分割技术将视频划分成一个个场景,在不同场景中嵌入不同的水印,从而使水印能够抵抗共谋攻击。在相同场景所有帧中嵌入相同的水印,通过冗余嵌入使水印对帧剪切具有鲁棒性。水印嵌入在各帧的DCT-SVD域,实验结果表明嵌入的水印不仅具有很好的不可感知性,而且对于低通滤波,中值滤波,高斯噪声以及H.264压缩具有良好的鲁棒性。水印的提取不需要原始视频信号,是一种盲水印算法;提出了一种在视频按照H.264标准进行压缩的同时嵌入水印的算法,其首先根据Ⅰ帧各个宏块的纹理特征以及所要嵌入水印的位数选择嵌入水印的宏块的位置以及水印扩频的倍数,将扩频后的水印序列嵌入在选择的宏块中。在宏块中嵌入水印时根据针对H.264压缩所特有的4×4DCT变换提出的人类视觉模型,选择合适位置的系数并且做适当的调整来嵌入水印。水印检测时只需要识别嵌入位置系数的符号即可以提取水印。实验证明,该算法兼顾了鲁棒性和透明性的要求,按照该算法嵌入的水印对重压缩,高斯噪声,椒盐噪声,高斯滤波攻击具有一定的鲁棒性。更多还原

【Abstract】 With the development of compression technology and network technology, people can easily and quickly copy and transmit a variety of digital video, and its copyright protection has become an urgent need to address the issue of digital video watermarking technology that provides an effective means to resolve. In this paper, the digital video watermarking has been studied mainly in the following areas have achieved some results:A blind adaptive video watermarking algorithm based on DCT(disc-rete cosine transformation) and SVD (singular value decomposition) is proposed. The algorithm first partitions the video stream into several shots, And both the frames of the shots are transformed into DCT-SVD domain. the watermarking sequence is embedded into the largest singular value. Experiments showed the new algorithm meet the transparency requirements at the same time meet the requirements of a certain robustness, resistance to low-pass filtering, median filter, noise and H.264 compression, and Collusion, Cut-and-splice attacks.The paper also propose a algorithm to watermark in video while it is being compressed in accordance with the H.264 standard.First,we choose appropriate macroblocks to be watermarked and proper multiple used for spread spectrum in accordance with various macroblock texture features as well as the length of the watermark sequence.The algorithm embed the spread spectrum watermark sequence into the choosed macroblocks. In accordance with 4×4 DCT transform human vision model proposed specific for H.264 compression,we select a suitable coefficient of each macroblock and make appropriate adjustment to embed watermark. Watermark is extracted by identifying the symbol of the coefficients of the selected locations. Experimental results show that the algorithm takes care of robustness and transparency requirements. The robust watermark embedded has good immunity to some attacks such as the recompression, Gaussian noise, salt and pepper noise, Gaussian filtering an so on.更多还原