基于视觉显著模型的音视频交叉水印算法

Audio and video cross watermarking algorithm based on visual saliency model

提出一种基于Itti视觉显著模型的音视频交叉水印算法。首先利用音频码流的低通幅值特性生成水印,并嵌入到视频码流中;其次解码视频码流的关键帧I帧,获取YUV序列上的亮度(Y)分量,并对其进行显著性分析;最后分块对非显著区域进行DCT变换,并通过量化索引调制嵌入水印。实验结果表明,该算法将嵌入水印后的关键帧平均PSRN值提高了1.678 5dB,缓解了水印不可见性和鲁棒性的矛盾,实现了音视频码流之间的交叉保护。

This paper proposed a simple audio and video cross watermarking algorithmbased on ltti＇s visual saliency model. Frist of all, it utilized the low-pass amplitude characteristic of audio strearningto generatewatermarking, embedded into the video streaming; Secondly, decoding the keyframe, I frame, of video streaming to obtain the Y component of YUV sequence, remarkably analyzing it; Finally, non-remarkable area is blocked into DCT, embedded intowatermarking by quantization index modulation. The experimental results show that the average PSNR of key frame, in this algorithm, is lifted by 1. 678 5 dB, which relieved the contradiction of watermark invisibility and robustnessand achieved the cross-protection between the audio and video streams.