基于光响应非均匀性的WhatsApp压缩视频来源识别

WhatsApp compressed video source camera identification based on photo response nonuniformity

光响应非均匀噪声(photo response nonuniformity,PRNU)是光学成像传感器成像时引入的一种独特噪声,可有效识别压缩视频的来源。针对现有算法提取压缩视频的PRNU效果并不显著的问题,论文提出了一种改进PRNU提取算法。首先,去除视频编解码的环路滤波器,对视频帧使用双密度双树复小波变换进行分解;然后对高频子带使用基于贝叶斯阈值估计的双变量收缩算法进行估计,再使用自适应加窗维纳滤波进行二次估计,得到噪声残差;最后用基于量化参数值加权的最大似然估计法聚合噪声残差,再与视频帧估计得到PRNU。实验结果表明:该文提出的方法在20 s时WhatsApp视频的识别率为75%。

Photo response nonuniformity(PRNU)noise is a unique noise introduced to optical imaging sensors during imaging and can be effectively applied to the source camera identification of compressed video.Due to the problem that existing algorithms do not produce significant effect on extracting PRNU of compressed video,an improved algorithm to extract PRNU was proposed.Firstly,the loop filter of video codec was removed,and the video frame was decomposed by double density-dual tree-complex wavelet transform.Then,the high frequency subband was estimated by bivariate shrinkage algorithm based on Bayesian threshold estimation,and the adaptive window Wiener filter was used for secondary estimation.Finally,after the noise residuals were obtained,they were aggregated by the maximum likelihood estimation method based on quantization parameter weighting,and the PRNU was estimated with video frames.Experiments on the VISION dataset show that the accuracy of the proposed PRNU extraction method in WhatsApp compressed video recognition is improved to 75%at 20 s.