利用运动矢量进行视频篡改检测

随着高质量的数字摄像机和功能复杂的视频编辑软件的出现,对视频进行恶意篡改已变得越来越容易;同时迅速庞大的视频数据也使得利用计算机自动检测恶意篡改的需求日益迫切.针对视频恶意篡改的特点并利用帧间内容的连续性,提出通过统计视频帧间预测的运动矢量,检测视频是否被恶意篡改.首先,提取各P帧的运动矢量幅度;然后利用滑动窗口、计算窗口中各运动矢量幅度与窗口中的幅度均值的距离.由于篡改点处运动矢量的幅度与所属窗口均值的偏移较大,从而能够有效检测出篡改及其位置.同时通过改变图像组大小进行再压缩,可以进一步降低漏检概率.实验表明所提方法能够有效地检测出视频在时间轴上的篡改.

Exposing Digital Forgeries in Video via Motion Vectors

With the advent of high-quality digital cameras and sophisticated video editing software,it is becoming increasingly easier to tamper digital video.Furthermore,rapidly enlarged video data has made the automatic detection of video tampering more emergent.According to the characteristics of tampering and by employing the contents continuity between frames,a scheme of tampering detection using statistics of motion vectors produced by inter-frame prediction is proposed.Firstly,the amplitude of motion vector of each P frame is extracted.Secondly,the distance between amplitude of each motion vector and mean of a sliding window covering a range of amplitudes is calculated.Due to the apparent drift from the window's mean at the tampered point,traces of tampering can be checked and positioned easily.Further,to reduce false rejection rate,re-compression is executed by altering the size of group of pictures.Experimental results demonstrate that the proposed scheme is able to expose video temporal tampering efficiently.