基于SIFT的Contourlet域抗攻击视频水印算法  

SIFT Based Video Watermarking Algorithm against Manifold Attacks in Contourlet Domain

作　　者：江烨倩 蔡明[1] 宋春霖[1] 

机构地区：[1]江南大学物联网工程学院,江苏无锡214122

出　　处：《计算机仿真》2018年第2期314-320,共7页Computer Simulation

基　　金：教育部留学回国人员启动基金(第48批)

摘　　要：针对现有视频水印算法抵抗旋转、缩放等几何攻击能力较差的问题,在水印不可见性较好的前提下,提出了一种基于尺度不变特征变换(SIFT)的轮廓波变换(Contourlet)域抗多种攻击的视频水印算法。先对关键帧做Contourlet变换,将低频部分时间方向分组进行一维离散余弦变换(DCT),将水印分别嵌入到各组DCT变换后的AC系数上,最后采用SIFT算子提取几何攻击前后视频帧的特征点,校正形变的帧从而提取出水印,并对各组提取的水印相同像素位置处的值求和确定最终水印。实验结果表明,上述算法在较好的不可见性下能有效抵抗旋转、缩放的几何攻击,且对噪声、滤波攻击、帧裁剪、帧删除、帧平均攻击也有较强的鲁棒性。The ability of resisting geometric attacks in existing video watermarking algorithms is poor. With the precondition of good invisibility, a video watermarking algorithm against the attacks in contourlet domain is proposed based on scale -invariant feature transformation （SIFT）. The key frames were decomposed using contourlet transfor- ming. The parts of low frequency were grouped and decomposed using one - dimension DCT. Watermark was embed- ded respectively in the groups of AC coefficients. The feature points were extracted with SIFT operator before and af- ter the attacks to correct geometric deformation. The summation of pixels in watermarks of each group determines the final watermark. The experimental results show that the algorithm can resist the rotation and scaling attacks effective- ly, and is of robustness to the attacks of noises, filtering, frame cropping, frame deleting and frame averaging with good invisibility.

关 键 词：尺度不变特征变换 轮廓波变换 视频水印 几何攻击 鲁棒性 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]