A tremendous amount of video data is transferred over the Internet from one location to another, and its amount is growing exponentially every day. Significant advances made in multipurpose video editing software technology have considerably increased the chances of digital video tampering/forgeries. Therefore, the authenticity of digital video has become important. In this paper, the passive algorithm based on the correlation consistency between entropy coded (DistrEn2D and MSE2D) frames for video forgery detection is proposed. The entropy-based texture feature, such as two-dimensional distribution entropy (DistrEn2D) and bi-dimensional multiscale entropy (MSE2D), is used in the proposed algorithm. This algorithm works in four stages and can investigate the presence of multiple forgeries in the videos. The first stage is pre-processing. In step second, the texture feature is extracted from the video frames. After that, inter-frame correlation consistency between Entropy coded frames is investigated to detect multiple forgeries. In the final stage, multiple forgeries are localized in the video using an abnormal point detection. An experimental result shows that the proposed algorithm (using DistrEn2D and MSE2D feature) provides better performance in identifying the forgeries present in the video.

大量的视频数据通过 Internet 从一个位置传输到另一个位置，并且其数量每天都呈指数增长。多用途视频编辑软件技术的重大进步大大增加了数字视频篡改/伪造的机会。因此，数字视频的真实性变得很重要。本文提出了一种基于熵编码(DistrEn2D和MSE2D)帧之间相关一致性的无源算法，用于视频伪造检测。该算法使用了基于熵的纹理特征，例如二维分布熵（DistrEn2D）和二维多尺度熵（MSE2D）。该算法分四个阶段工作，可以调查视频中是否存在多个伪造。第一阶段是预处理。第二步，从视频帧中提取纹理特征。之后，研究熵编码帧之间的帧间相关一致性以检测多个伪造。在最后阶段，使用异常点检测在视频中定位多个伪造。实验结果表明，所提出的算法（使用 DistrEn2D 和 MSE2D 特征）在识别视频中存在的伪造方面提供了更好的性能。