本文提出一种在原始视频DCT(离散余弦变换)与SVD(奇异值分解)域自适应嵌入水印的算法。算法首先将视频流分割成一个个场景,场景中的视频图像被变换到DCT-SVD域中,水印量化嵌入在最大奇异上,实现了水印盲检测。同时实验证明该算法在满足透明性要求下也满足一定的鲁棒性要求,能够抵抗低通滤波、中值滤波、椒盐噪声、高斯噪声、H.264压缩攻击以及统计攻击和帧剪切等攻击。-This paper presents an original video in the DCT (Discrete Cosine Transform) and SVD (singular value decomposition) algorithm for adaptive watermark embedding. Firstly, a video stream is divided into scenes, scenes of the video image is converted to the DCT-SVD domain, the watermark embedded in the largest singular quantification, the realization of the watermark signal. Furthermore, the experiments show that the algorithm meet the transparency requirements are met under certain robustness requirements, can resist the low-pass filtering, median filtering, salt and pepper noise, Gaussian noise, H.264 compression attacks and attacks and frame shear statistics attack.