A fast and effective preprocessing method for video copy detection

This paper proposed a fast and effective video preprocessing method to detect geometry distortion applied on original videos. If a video copy detection method is effective, it needs to be robust to the many digitization and encoding process that give rise to several distortions, including adding boards, rotation and so on. Most of the video copy detection algorithms proposed so far focus on coping with take all possible kind of distortions into consider and design a robust video fingerprint, however, design such kind video fingerprint is not a easy work. We proposed a method to be robust in detecting and adjust the geometry distortions. To this end, we get a sequence of frames, and extract the video region. Then by detecting the quadrilateral region which contains video, we transform the distorted video back to original state. The proposed method has been extensively tested and the results show the proposed scheme is fast and effective in dealing with several types of geometry distortions.