Abstract: An MPEG video hierarchical content watermarking authentication scheme that utilized the local and global features among video frames was proposed to find out whether the content of video had been maliciously changed or not. The two employed kinds of features in the scheme, the global one and the local one, are semifragile, which can tolerate the legal content persistent operation and sensitive to the tamper operations. The global feature was abstracted from the relation coefficients of DC coefficient matrix in the I type pictures, which could detect the content tamper with a lower computation payload. After the tamper frame was decided, the tamper positions could be further localized with the local features. Experimental results show the satisfying performance of the proposed scheme. The fit choosing strategy in this scheme can reduce the computing payload about 20 percent. In a word, the content feature based video authentication scheme can lighten the way to build practical video authentication methods.

Key words: video processing, semi-fragile watermarking, Least Significant Bit (LSB), video compression coding

摘要： 为对MPEG视频是否经受恶意篡改实施有效认证，提出一种融合全局内容特征和局部内容特征的MPEG视频分层CFA认证方案。该方案选取的特征具有两类“半脆弱”的特征，即全局特征和局部特征。全局特征由I帧分块DC系数矩阵的相关系数构建，而局部特征则反映了视频宏块的基本信息。基于两类特征的构建，使得方案对于可接受的“内容保持”处理具有鲁棒性，而对恶意篡改具有敏感性，从而增强了对内容篡改行为的识别能力。同时，方案能够以较小的计算复杂度对视频内容是否遭受恶意篡改进行快速判决，平均运算时间减少了约20％。

关键词: 视频处理, 半脆弱水印, 最低有效位, 视频压缩编码