基于改进的JND模型的视频水印算法

Video Watermarking Algorithm Based on an Improved JND Model

针对视频版权的保护,提出了一种基于改进的JND模型的视频水印算法。此算法首先选取二值图像作为图像水印,利用Logistic混沌映射对图像进行加密;然后对载体视频进行关键帧提取,将提取的帧进行4层小波分解,结合分形理论对小波系数进行D树和R树的存储,并根据人类视觉系统(HVS)提出改进的JND模型,进而将处理后的水印信息按照三个方向自适应的嵌入到D树和R树中。实验结果表明:本工作的算法具有较好的不可见性,且针对噪声、高斯滤波、旋转等攻击具有良好的鲁棒性。

For the protection of video copyright,this paper proposed a video algorithm based on an improved JND model.This algorithm firstly selects two-binary image as the watermark image,encoding the image by the logistic chaotic mapping;then extracting the key frame from the carrier video and decomposing the frame by four-layer wavlet,divide the wavlet coefficents and store them in the D-tree and the R-tree combined with the fractal theory,we proposed an improved JND model based on HVS and embedding the processed watermark information into the D-tree and the R-tree adaptively according to three directions.The experimental results show that this algorithm has a good invisibility,and the attack on Sharpening,Gauss filtering,noise,frame shear has good robustness.