基于视觉掩蔽模型的视频水印算法

A video watermarking algorithm based on visual masking model

提出一种基于Watson视觉掩蔽模型的MPEG-2视频水印算法。将水印图像和水印属性转成二进制比特流,作为嵌入视频的水印信息。计算视频关键帧的DCT掩蔽阈值,根据阈值分布特点设计动态的水印嵌入量化步长,调整水印强度,采用量化索引调制的方式嵌入水印信息,保证嵌入的水印对视频的影响在视觉失真范围内。实验结果表明,该算法将嵌入水印后的关键帧平均PSNR提高了0.78dB,有效缓解了水印不可见性和鲁棒性的矛盾。

This paper proposed a video watermarking algorithm of MPEG-2 based on Watson＇s visual masking model. It transforms the watermark image and its attributes into binary bit stream working as actual watermark embedded into the video. Calculate the DCT masking thresholds of video key frames and set dynamic quantization steps for watermark embedding to adjust watermark strength according to the distribution of thresholds. Adopt the method of Quantization Index Modulation to embed watermark and ensure that the impact of embedding watermark on the original video within the range of visual distortion. The experimental results showed that the watermark invisibility is lifted by 0. 78 dB in average PSNR of key frames and effectively relieve the contradiction between watermark invisibility and robustness.