Detecting special frames in sports videos

Abstract

Statistically, among the common video types, sports videos have more interesting moments than other types of videos. And it is often desired to automatically detect some special frames in sports videos, like close shot of athletes, illegality penalty, controversial goal, game result judgement, etc. In this thesis, we focus on two scenarios of video frame detection in sports videos: transition frame detection and video frame detection for sprint judgement.

For the first scenario, we study a particular type of video transition, i.e. dissolve. Dissolve is the basic artificial effect of gradual transitions in video sequences, which is widely used in TV programs and movies. Through dissolve detection, interesting and important video temporal segments can be easily located and indexed for various applications. In this research, we present a simple yet effective dissolve detection algorithm, which takes into account the dissolve properties in both temporal and spatial domains. Specifically, in the temporal domain, we use frame difference to capture the dissolve characteristics. In the spatial domain, the central area is given bigger weight than the four sides. Experimental results show the good performance and low computational complexity of our proposed algorithm.