Construction And Performance Evaluation Of Video Forgery Detection Database

Multimedia technology is playing an increasingly important role in our daily life.As an efficient way for transmitting information,video has become the major approach of disseminating information and has been widely used in news reports,social communication and forensic evidences.With the development of various multimedia editing software,video tampering has become easy which damages the credibility of digital videos.In this context,digital multimedia forensics technologies have emerged as one of the important branch of digital forensics.The research on video forgery detection,which is used to determine the originality and authenticity of videos,has obtained a lot of attention.A variety of video forgery detection algorithms have been proposed,either inter-frame or intra-frame video forgery detection algorithms.However,there is a lack of standard video database in the field of video tampering detection.The main contributions of this thesis are listed as follows:1.We have created a Video Forgery Detection Database(VFDD)that could be used as a third party public database for the research of video forgery detection.The database currently comprises 640 video clips,including 505 clips of original videos captured by various imaging devices and tampered videos created with inter-frame(57 clips)as well as intra-frame(78 clips)manipulations.We introduced the details of the construction of the database,such as the capturing devices,shooting conditions,video formats.We also described the process of rendering tampered videos with inter-frame and intra-frame manipulation.2.We have conducted introduced and implemented eight typical algorithms for video forgery detection.Among the algorithms,six are for the detection of inter-frame forgery while the other two are for the detection of intra-frame forgery.The algorithms were implemented with OpenCV and Matlab and tested on the VFDD database.We analyized the results of each algorithm and have provided ideas for improving the algorithms.3.Using the eight-typical video tampering detection algorithms,we have conducted thorough experiments to evaluate the effectiveness of the VFDD database.The algorithms were tested with optimal thresholds.Based on the sensitivity and specificity measurements,we have also listed the various factors that would affect the accuracy of detection.It has been shown that the VFDD database can effectively demonstrate the advantages and disadvantages of different algorithms.The VFDD database is proved to be an effective evaluation database for various video forgery detection algorithms.