Video Inter-frame Forgery Detection Based On Continuity Feature

With the widely applying of low cost digital devices(such as digital cameras and video cameras), and the popularizing of video sharing websites(like Youku and YouTube),digital video is becoming increasingly important in our daily lives. However, digital video can be easily altered by the advanced video edition software, which makes its authenticity can’t be guaranteed anymore. Especially when these videos are used as evidence in court, it will seriously hamper the court to make a fair judgment if the evidence can’t be authenticated. Video inter-frame tampering, such as frame deletion and frame duplication, is one of very common operations of forgery, in this paper, we propose two video inter-frame forgery detection algorithms based on uncompressed domain and compressed domain respectively.From the perspective of the non-compressed domain, we presents an inter-frame forgery detection algorithm based on the consistency of velocity ?eld. The main principle of the approach is that any tampering operation will introduce breakpoints in the velocity ?eld intensity(VFI) sequence of original video whose VFI sequence is considered as continuous. And the different types of tampering operation will introduce corresponding number of discontinuity points. According to this principle, the algorithm can effectively distinguish the tampered video, identify the forgery types(i.e., frame deletion and frame duplication) and locate the manipulated positions in forged videos as well. A video database is established by ourselves to test the algorithm, experimental results show the effectiveness of our algorithm. Finally, comparative experiment is conducted with some analysis.From the perspective of the compressed domain, we presents the other frame video deletion detection algorithm based on macro-block feature. The main principle of the method is that frame deletion tampering makes the number of I-MB and MV-MB increasing and the number of ZMV-MB decreasing in P-frame, which originally is a I-frame in single compressed video. Therefore, we can distinguish the forgery video by detecting the periodic changes of different types of macro blocks. Video database contains 432 MPEG-2 video clips compressed with different GOP size and quanti?cation scales. Each video is manipulated by randomly deleting a frame for 10 times to get the average detection accuracy. The experimental results demonstrate the effectiveness of the algorithm.