基于运动目标子块分割的视频零水印算法

VIDEO ZERO-WATERMARKING ALGORITHM BASED ON SEGMENTATION OF SUB-BLOCK OF MOVING OBJECT

为了提高视频水印的鲁棒性和隐蔽性,提出一种将水印与视频内容特征相结合的零水印方案。通过帧间欧氏距离选取关键帧,利用改进帧差法提取视频关键帧中运动目标,对关键帧的运动目标进行子块分割,选择运动目标几何形心所在块构成三维体块;再将三维体块分割成多个大小相等的小三维体块,利用小三维体块中所有子块的离散余弦变换直流系数构造特征向量;最后利用特征向量实现将水印零嵌入到视频内容中。实验表明,该算法未破坏原视频内容,不仅具有很好的隐蔽性,还能有效地抵抗旋转、压缩编码等攻击。

In order to improve the robustness and invisibility of video watermarking,we propose a zero-watermarking scheme which com-bines the watermarking with video content characteristics.It selects the key frames through interframe Euclid distance,uses improved frame difference algorithm to extract the moving objects in key frame,and segments the moving objects of key frame into sub-blocks.After selecting the blocks at the geometric centroid of moving object to reconstruct three-dimension block,the scheme then segments the three-dimension block to a couple of small three-dimension blocks in same size,and constructs the feature vector using DC coefficients of discrete cosine trans-form of all blocks in small three-dimension block.Finally,it uses feature vector to realise embedding the zero-watermarking into video con-tents.Experiments show that the algorithm does not damage original video content,apart from having very good imperceptibility,it can also effectively resist the attacks of rotation and compression coding,etc.