The Composition And Comparison Of Sports Video

In recent years, digital video processing technology is introduced in sports training.This kind of technical innovations in training methods greatly promote the efficiency of the sports training. However, the reseach involves video object segmentation, human motion tracking, video content understanding, etc. In the fields of video processing and computer vision, the researchs on human motion analysis, human body segmentation and object based video matting are attracting many researchers' attention with thire wide application prospect. Because of the complexity of human motion and background variation, the existing research methods impose limitation on complex human action and dynamic background.A new approach to compose the diving videos based on human body segmentation and obtaining motion region information is brought forward in this paper. Instead of aiming at a static background, it can deal with diving video with dynamic background. The main scheme is to obtain motion region information of the humanbody from the diving video then use this information to matting the moving humanbody object into the other video. The image difference based segmentation, global motion estimation (compensation) and matting are used. The final result is given to show that our method does well in keeping the definition of the athlete's body obtained from the original video. The main contributions are as follows:1. Accurate initial value and outlier filtering threshold selection in GMEGlobal motion estimation (GME) is to estimate the parameters of camer motion that causes the movement of background in video. The pixel correspondences between adjacent frames can beobtained by GME. Thus, the dynamic background in diving can be simplified to a static one with GMC (global motion compensation). An accurate initial value selecting method is proposed to improve Konrad's GME method in MPEG-4 validation model in order to avoid getting trapped into a wrong local minimum. A threshold selection from gray level histogram using Otsu method is used in GME to filter the outlier during reiteration. By filtering the outliers, the noise is eliminated effectively in parameters estimation and the iterate times is greatly decreased. Compared with Konrad's method, the method improves the speed of GME nearly 2 times.2. Video matting using motion region information with keeping the definition of moving objectAn automatic composition algorithm for sports video in same scene is proposed. The information of the foreground areas in current frame is obtained with GME and GMC between current frame and temporally adjacent frames. The composition frame is constructed based on the information of the foreground areas of both pre-composition frames using ALPHA matting. Experiments demonstrate that proposed method is robust to the dynamic scene and able to compose the sports video keeping both of the foreground areas of pre-composition frames clear.3. Implement the Composition and Comparison of Sports Video in DVCOACH systemWith all the methods mentioned aboue, this paper implement the composition and comparisonfunction in DVCOACH system.