H.264视频水印技术的研究与实现-智能与分布计算实验室

H.264视频水印技术的研究与实现

姓名	倪丕祥
论文答辩日期	2007.05.31
论文提交日期	2007.06.10
论文级别	硕士
中文题名	H.264视频水印技术的研究与实现
英文题名	The Research and Implementation of Video Watermark technology base on H.264 standard
导师1	李瑞轩
导师2
中文关键词	数字水印;H.264编码标准;离散余弦变换;差分能量算法
英文关键词	Digital watermark;H.264;Discrete Cosine Transform;Differential Energy Watermark
中文文摘	H.264作为一种新型的基于对象的压缩标准，在视像电话、视像电子邮件和电子新闻等方面有着广泛的应用前景。随着计算机网络和多媒体(数字图像、视频、音频)技术的广泛应用，多媒体知识版权保护成为一个迫切需要解决的问题。视频水印技术作为多媒体版权保护的重要手段得到了越来越广泛的研究与应用。基于H.264的视频数字水印版权保护技术也由此受到了人们的广泛关注。着重从以下几方面进行了深入研究和探讨。首先，全面分析了H.264视频压缩标准的原理，论述了视频水印技术与视频压缩标准相结合的必然性。接着，结合H.264压缩标准，讨论了在压缩标准中水印作用域的选择问题。分析了现有的基于DCT域的视频水印算法，重点介绍了经典的差分能量算法，并分析了该算法的优点与不足。在此基础上设计了一种基于DCT域的H.264视频水印技术方案，并详细给出了有关算法设计实现的过程。在该方法中，将水印信息经加密后嵌入到视频I帧的亮度DCT交流系数(AC)中，水印嵌入和提取时不需要完全解码，大大减少了运算的复杂度，提高了实时性。水印检测时不需要原始视频，实现了盲检测，在节省了时间的同时取得了良好的视频效果。另外由于把水印嵌入到亮度DCT的交流系数中，可在保证视频效果不失真的前提下，使水印具有很强的健壮性。此外，在嵌入水印之前，还对水印信息进行了若干预处理，进一步保证了算法的鲁棒性；对视频也做了诸如宏块置乱等处理，提高了算法的不可见性。试验结果表明，水印嵌入前后，视频质量没有明显的变化，在无攻击时检测到的水印信息和原水印信息相比也没有明显的变化。方案中的水印只嵌入在视频I帧中，不修改P帧和B帧，对帧跳跃和帧删除攻击稳健，因为I帧不可以被跳跃或删除，且对一般的图像处理如剪切、滤波、锐化和加噪声等，都具有较好的鲁棒性。
英文文摘	As a new type of compression standard based on object, H.264 can be widely used in the fields of video phone, video email, electronic news, etc. As one of the important means of multimedia copyright protection the watermark technology of the video has got more and more extensive research. So, digital watermarking copyright protection technology based on H.264 becomes many people’s focus in recent years. This paper introduces analyze the H.264 principle at first. After that the inevitability which video watermark technology and video frequency compression standard combined is discussed. Combining the compress standard of H.264, this paper has discussed the choice question of watermark embedded domain in H.264 and analyzed the existing video watermark algorithm based on DCT domain. Based which, a method of video watermark which is based on DCT domain in H.264 is designed. And then this paper provides the design and implementation course of algorithm. The watermark is retrieved not to need primitive video. The blind retrieving is achieved. In addition, the watermark has strong robust when the video frequency effect is not dropped, because the watermark is embedded in luminance AC domain. Experimental results show that video quality has not visible change between in front watermark embedded and at the back. And the watermark information compared with original watermark has not obvious change. In this method, the watermark is only embedded in I frame not in P frame and B frame. This method is resisted the skip and deletion of the frame in effect because the I frame is not skipped or deleted. And it has strong robust under the image processing attacks such as cutting, filter, sharp and yawp etc.