同源视频Copy-Move篡改检测及恢复①

摘要

With regard to the copy-move forgery of video frame sequence, this paper proposes a algorithm for video detection and recovery base on the correlation of the contents between frames. Firstly, the video frames are translated into a series of consecutive image frames, and by blocking the frame to extract eight dimensional feature vectors. With Euclidean distance computed the correlation between frames, a deviation matrix is import in the algorithm to reduce the false alarm rate. A threshold value is used to localize the modifications to a frame level and recover the original video. Experimental results show that this method can effectively detect the copy-move forgery of video frame sequence and video restoration.%针对同源视频序列的copy-move篡改方式，提出一种通过度量图像内容间的相关性，来实现对视频序列的copy-move篡改检测并恢复的方法。首先将视频帧内容转化为一系列连续的图像帧，对图像分块，提取每帧图像的8个特征矢量，再利用欧氏距离计算帧间相关性，并通过添加偏差矩阵构造动态偏差阈值，检测出copy-move 篡改序列且精确至帧，从而实现对视频序列的篡改检测与恢复。实验表明，该算法对同源视频序列的copy-move 篡改检测及恢复能够取得理想的效果。