Detection of inter-frame forgeries in videos

Video forgeries are used to hide an event or object, or to mislead the viewer by altering the flow of the video. Among video forgeries, inter-frame forgeries are among the most common types of forgeries. In this thesis, the problems of detecting inter-frame forgeries in videos are investigated and new methods are proposed to solve them. The studies were tested on both the existing database and the new database created within the scope of the thesis. Common databases commonly used in the literature and specially prepared realistic video examples were used. Firstly, texture and edge detection methods are used to detect repetition forgery in uncompressed videos. Secondly, we propose a frequency-based approach to detect duplication forgery that works on compressed videos. The proposed method was tested against blurring attacks and successful results were obtained. Thirdly, approaches to detect multiple forgeries have been proposed. For this purpose, we designed a similarity-based approach for the detection of deletion forgery and a hybrid approach for the resolution of all inter-frame forgeries Hybrid operation is made resistant to compression attacks.