Video summarization based on user interaction

Abstract

It is frequently the case that viewers want to watch sports in less time than a game takes to play. Computing a summarization of a game depends upon information about each play. In contrast with previous video summarization work this paper describes how the interactive behavior of prior viewers can be used to compute which plays are most interesting. This passive feedback is examined as a possible source for the degree of interest in a play. In addition, a mechanism is described for briefly showing the game action that was removed to create the summary. This helps to preserve the continuity of the game in the viewer's mind.