Design And Implementation Of Video Forgeries Detection System Based On Content

With the popularity of video editing software and operation of civilians, the tampered videos often occur. The probability of people facing the tampered video increases greatly. So, the technologies of ensuring the integrity and authenticity of the digital video have become the urgent needs in the today’s society, especially the security of digital image and video.In this thesis, a video tamper detection algorithm is proposed to detect the deletion, insertion and copy in the time domain based on the texture spectrum. The main work of the thesis includes:(1)The advantages and disadvantages are summary based on the analysis of the video tamper detection technologies from the three facts, that’s the shooting device noise, coding characteristics and visual content.(2) A passive tamper detection algorithm is proposed based on video texture spectrum. The technology of the image texture spectral representation is introduced into this field for the first time. Secondly, the similarity model is established between the inter frame texture visual content. Thirdly, the anomaly degree model is put forward based on the similarity sequence by using LOF algorithm. Finally, the abnormal position are detected by using the threshold method. The simulation results demonstrate the effectiveness of the proposed algorithm.(3) A video tamper detection system is design with a friendly interface based on the Matlab software platform and GUI development environment. The system operates simply and implemented without the professional knowledge. The testing results show that the system designed in this paper can detect the video frame deletion, insertion, copy, and paste effectively. The accuracy of three tamper detection all exceed 95%, and recall rate exceeds 90%. The time of the average processing for each frame is less than 0.5 seconds.