基于多特征的半脆弱视频水印算法研究

Research on semi-fragile video watermarking based on multi-feature

针对目前视频常遭到恶意篡改的问题,提出了一种半脆弱视频水印算法,可以实现视频内容完整性认证并能准确定位篡改区域。该算法通过把视频I帧分成不同区域Ia,Ib来分别提取特征和嵌入水印,避免了水印嵌入对特征的影响;利用边缘轮廓和块间DCT低频系数能量关系不变的特点作为特征生成水印,能更好地代表原视频帧内容。然后,置乱水印并嵌入到视频帧中,通过比较重新提取的水印和特征完成视频内容认证。实验结果表明,嵌入的水印具有较好的不可见性,在对视频内容进行认证时能更准确地定位被篡改的区域,具有较高的查准率和查全率。

Aiming at the problem of video being malicious tampered, this paper proposed a semi-fragile video watermarking algorithm that can certify video content integrity and locate the tampered region accurately. The algorithm divide the video frame into two regions （Ⅰa. Ⅰb）, where the watermarking is embedded and the features are extracted respectively. It can avoid the influences on the features. The characteristic that the energy of DCT coefficient and the edge can keep same, that can be treated as features for generating watermarking. So the generated watermarking can represent more full contents of the original video frame. This paper disorder the watermarking and embed it into the video frame, then re-extract the feature and the watermarking to authenticate the video content. Experimental results show that the embedded watermark has good invisibility, video content can be authenticated more accurately with high precision and recall.