Due to vast amount of archived video content around the world, flexible and efficient soccer video summarization systems are required. In this paper, we present an efficient yet accurate framework for soccer video summarization in encoded MPEG videos. A novel logo transition detection method based on spatio-temporal template is proposed. Semantic features along with replays and audio activity are exploited for summary generation based on user defined parameters and rank based video summarization scheme. We conducted objective and subjective evaluations on ten hours of broadcast soccer videos. Rank correlation measures are used for objective video summary evaluation. Our system achieved satisfactory accuracy and speed of nearly two times faster than real-time.

Keywords