摘要
针对现有视频水印算法在抗几何攻击能力方面的一些不足,提出一种基于关键帧和小波变换的盲视频水印算法。依据视频帧间差欧氏距离提取视频关键帧并分块,结合人类视觉系统特性,从关键帧不同的块中选择亮度敏感性和纹理敏感性最高的块构成三维体块,进行三维小波变换,将置乱的水印图像以不同的嵌入强度嵌入到小波系数的低频区域,利用阈值门限实现水印的盲提取。结果表明,该算法提高了水印嵌入容量,对于针对视频水印的攻击具有较好的不可见性和稳健性。
A blind video watermarking algorithm based on key frames and 3D wavelet is proposed in this paper in order to improve the ability of the watermark to resist geometric attacks. The key frames are selected based on Euclid distance of Video frame difference, the key frames are segmented to blocks, and the key blocks are selected based on the video characteristics such as luminance sensitivity and texture sensitivity. Then ,the key blocks are reconstructed to three dimension blocks and the scrambled watermark is adaptively embedded into the low frequency domain of the three dimension wavelet decomposed coefficients of the blocks. The watermark is retrieved blindly based on JND. Results demonstrate that the algorithm improves the watermark embedding capacity,and obtain a good imperceptible and robustness to video attacks.
出处
《电视技术》
北大核心
2013年第9期12-15,33,共5页
Video Engineering
基金
江西省教育厅科技项目(GJJ12345
GJJ12349)
关键词
视频水印
关键帧
人类视觉系统
视频分块
video watermarking
key frames
human visual system
video segmentation
