Detection Of Blue Screen Matting And Compositing In Videos Based On DCT Statistics

In recent years, with the development and popularity of computer techniques, image and video editing software with complete function and easy operation have received more and more attention. Therefore, simple editing of digital images and videos, which is indispensable in everyday work, life and entertainment, has transformed from specialized skills of professional to individual behavior of demanders and fans. However, people start to doubt the authenticity of the contents of images and videos which they firmly believed. If tampered videos are used in improper ways, most adverse effects will be caused. Currently, there is no sound supervisory system and detection techniques perfect enough, which has been a key problem to ensure a steady development of information security field.Blue screen matting and compositing of digital videos is a common technique in special effects production of videos, whose detection can effectively ensure the integrity and authenticity of video contents. In this thesis, an algorithm to detect blue screen matting and compositing based on DCT coefficients statistics features is proposed. In view of different sources of foreground and background in composited videos, quantized DCT coefficients of foreground and background are extracted separately according to different quantization scale factors. In order to analyze the differences between the two distributions of coefficients, radian distance is introduced as a quantitative description, which is also the criterion to make the final judgment. In this thesis, the process of blue screen matting and compositing are implemented in the software of Adobe Premiere Pro and various bitrates are set for the composited videos for testing. Experimental results have shown that the algorithm proposed can manage to detect composited videos in different bitrates whose background videos are from diverse encoders.