Video analysis and shot-boundary detection

Abstract

Many multimedia applications demand retrieval of video data. For example, a coach of basketball team analysis how does another team organize their offense by replaying the video clip which contains the offense. Due to the large archives of video data, it becomes necessary to automatically extract the video contents and index the abstraction of video data. To achieve these goals, it is necessary to analyze the video sequence and segment it into individual shots. In this project, soccer videos are used for analysis. 4 types of shots are defined and the corresponding shot boundary is detected whenever a frame is found that belongs to different shot type with the previous frame, where frame is a still picture and video are consists of frames. The 4 types of shots are long shot, in-field medium shot, close-up shot and out of field shot. The testing data are MPEG clips. Color histogram is the main information to analysis videos, and shot boundary is detected by comparing information about color histogram of different frames in my project.