The Research Of Several Key Technologies For Content-Based Video Retrieval And Search

In this paper, we propose a framework for highlights extraction based on cinematic features and cinematic template. It is full automatic, generic, and effective. It includes some novel low-level algorithms, such as shot boundary detection, shot classification, slow-motion detection algorithms, and high-level algorithms, the goal detection algorithms based on the cinematic template. Video goes through the shot boundary detection module first. Then, the shots are put into the shot classification module and slow-motion detection module separately. At last, if the shot type and the slow-motion shot satisfy the rules defined by the cinematic template, the shot is the goal shot. According to the feature of soccer video, we introduce some new features, and the algorithms get better results than that of traditional ones.