Detecting Forgery From Video With Pattern Noise Consistency

With the advent of the information age and the rapid development of computernetwork technology, multimedia information has been popular with us in daily life,and videos have become a very important source of information. But as the popularityof video editor, it is becoming increasingly easier to tamper digital video, which havecaused serious affection. Therefore, in order to ensure the authenticity and integrity ofvideos, the tamper detection has been a research hotspot in video forensicstechnology.In this thesis, the video forgery detection scheme based on consistent patternnoise is proposed, which aims to detect the tempering of video that moving objectsmay be artificially removed in static background and motive background, and thosevideos are always tempered by Adobe Photoshop. Based on the analysis of digitalvideo forensics, the temper localization can be detected by the extraction of the sensorlegacy pattern noise, and then the reference model of pattern noise can be built.According to the consistency of the pattern noise, the relationship between blockregion in target’s frame and its corresponding region in reference model can be usedas a basis of whether the video has been tempered. At the same time, for the motivebackground, an adjacent frame block matching method can improved the accuracy. Atthe end of this thesis, effectiveness and reliability of the algorithm are proved byMATLAB simulation, and then the simulation results are also analysised.