基于NSCT域核模糊聚类的视频水印算法

Video watermarking algorithm based on fuzzy kernel clustering in NSCT domain

为减小水印嵌入对视频视觉效果的影响,结合人类视觉系统,提出了一种基于无下采样Contourlet变换(nonsubsampled contourlet transform,NSCT)域核模糊聚类的视频水印算法。首先利用独立分量分析提取视频同一镜头中视频序列的静态分量;然后根据人类视觉系统的掩蔽特性和图像的局部特性,对静态分量NSCT后的带通子带进行核模糊聚类分析;最后将置乱后的水印嵌入到适合嵌入水印类的子带系数中。文中给出了视频嵌入水印后的平均峰值信噪比、算法鲁棒性比较和视频被攻击后所提取的水印。实验结果证明该算法能取得极佳的视觉效果,同时还能抵抗空间域和时间域的常见攻击。

To reduce the effect of embedding watermark into a video on the visual quality of the video,combined with human visual system,a watermarking algorithm based on fuzzy kernel clustering in NSCT domain is proposed.First,the static component of a video segment in the video shot is extracted using independent component analysis.Then the static component is decomposed with NSCT and the bandpass subbands are analyzed after NSCT processing using fuzzy kernel clustering according to the imperceptible characteristic of human visual system and the image local characteristic.Finally,a scrambled watermark is embedded into the coefficients of the bandpass subbands suitable for watermark embedding.The peak signal to noise ratio（PSNR） of the video embedded with watermark is evaluated.The robustness of the algorithms is also evaluated.And the watermark extracted from attacked video is given from experiment result.Experimental results show that the proposed algorithm can preserve high visual quality of video and the watermark has enough robustness to resist common attacks in spatial domain and temporal domain.