基于小波变换和神经网络的视频水印算法

Video Watermarking Algorithm Based on Wavelet Transform and Neural Network

针对视频版权的保护提出了一种基于三维小波变换和RBF神经网络的视频水印算法。本算法采用二值图像作为水印,对水印图像进行随机置乱预处理以增强其安全性能;对宿主视频进行视频分割,并分别对得到的各序列做三维DWT变换;把水印嵌入到经过三维变换的低频子带中,在水印提取过程中,结合RBF神经网络模型,而不需要原始图像,快速有效提取出水印。实验结果表明,该算法具有较强的不可见性,且针对帧压缩、旋转、滤波和比例缩放等具有良好的鲁棒性。

For video copyright protection,this paper proposes a video watermarking algorithm based on three-dimensional wavelet transform and RBF neural network.This algorithm uses binary image as a watermark,first of all,to enhance its security performance,with the random scrambling watermark image pre-processing;then segments the host video and separately transform each sequence using three-dimensional DWT;embeds the watermark into the low-frequency sub-band.In the watermark extraction process,combining with RBF neural network model,the watermark can be quickly and efficiently extracted without the original image,the experimental results show that this algorithm has a strong non-visibility,and has high robustness for the frame compression,rotation,scaling and filtering.