基于小波变换域的运动目标检测视频水印算法

针对现有视频水印算法稳健性不足的问题,结合人类视觉系统,提出一种基于小波变换域的运动目标检测视频水印算法。该算法首先将原始帧图像进行8×8分块,分别对每一子块进行Daubechies复杂小波分解,采用基于近似中值滤波的帧间差分法,有效地解决阴影、重影、噪声等问题,更加准确地检测运动目标并标记运动区域；然后对水印进行Arnold置乱和混沌加密预处理,来加强水印的安全性；最后将预处理后的水印自适应嵌入在运动区域和高纹理复杂区域中。实验结果表明,该算法具有较好的不可感知性,并对常见的视频攻击具有较好的稳健性。与现有三种算法的性能比较,说明了本算法的优越性。

Video watermarking algorithm based on moving object detection in wavelet domain

For the robustness deficiency of the existing video watermarking algorithms, combined with the human visual system,a new video watermarking algorithm of moving object detection based on wavelet transform domain was put forward. Firstly, the original image was divided into 8 x 8 sub-blocks and each sub-block was decomposed by Daubechies complex wavelet. In order to solve the problems of ghosts, shadows and noise, frame difference method based on approximate median filter was used to detect the moving targets and mark the movement area more accurately. Then the watermarking was preprocessed by using Arnold scrambling and chaotic encryption to enhance the security of the watermark. Finally the moving area and the area with highly texture complexity were chosen as the embedding area of video watermarking. The experimental results show that the proposed algorithm has the property of imperceptibility and good robustness against the common video attack The performance comparison with three existing algorithms also reflects the superiority of the proposed algorithm.