Football Video Slow Motion Detection And Video Summary Generation

As the fast development of media technique and Internet, needs of media information are wanted more and more. Sports video is an important part of video, and soccer video is popular for sport fans. But traditional video abstract are made manually, which requires great time, repetition and energy. So automatic extraction is hot for research of video processing.This paper focuses on slow shot in video and its function for audience, and presents a video abstract system based on slow shot.In shot detection module, based on presented algorithms, this paper offers an automatic shot detection method based on pixel histogram in compressed video data. The method uses slipping window associating with algorithm based on histogram to fix on a threshold dynamically. The method also introduces the relationship between similarity in the same shot and difference between different shots for shot detection. The method this paper presents can fix on a optimal threshold according to the video statistical information, and detects shots in the video.In slow shot detection module, based on general zero-crossing method, this paper presents a new zero-crossing algorithm which introduces slipping window most used in shot detection in order to fix on threshold automatically. The new method can get good result in different videos.In abstract extracting module, the system uses the function which DShow offers to users to make the slow shots detecting previously abstract. This paper also presents a abstract extracting frame based on ontology. The steps are as followed. Firstly, key frames are extracting from the video, then distinguishing whether the key frames are shoot frames based on ontology. If a key frame is about shoot, the slow shot which is corresponding to the key frame is the abstract needing.Experiment result shows both the shot detection algorithm and the slow shot detection algorithm run fast with high accuracy. But the algorithms have some slow shot missing and fault, which needs improvement in the future.