Query Events in Soccer Video Using On-Screen Texts

On-screen text is a rich information resource to query events in soccer video due to its close relation to what happens on the screen. However, low resolution, clutter background, unknown font, size, color, etc. prevent the efforts of using this resource for querying. This paper presents a novel approach for querying events in soccer video using on-screen texts. The proposed approach is completely automatic and independent to languages since it recommends the users to query events by keywords in image-form which are agents of clusters of stationary on-screen textboxes which are localized and extracted properly by a novel mechanism before being indexed and clustered quickly by an efficient variant of HAC algorithm. Our system is experimented and verified on 17-hour SUZUKI AFF Cup Tournament 2008 video data set. It shows the encouraging results in querying soccer events in video only using image-form keywords.