AUTOMATIC VIDEO SUMMARIZATION BY DISCOVERING FREQUENT PATTERNS WITH SUPPORT VECTOR CLUSTERING

摘要

In this paper, we present a novel approach for automatic video summarization by discovering frequent shot patterns with support vector clustering. The frequent shot patterns in video are defined as a sequence of video shots that occur frequently within a time interval. First, support vector clustering is implemented to cluster similar video shots into groups based on the shot boundary detection. Then frequent shot patterns are extracted in order to remove the visual-content redundancy among video content clusters and generate the video summary. The experiments show that support vector clustering combined with frequent shot patterns discovery is an effective approach for video summarization.