An efficient technique for summarizing videos using visual contents

We can summarize a video using its 'important' and/or 'interesting' scenes. The scenes selected, however, depend on the purpose of the summary. A good summarizing technique should be able to support various summarization needs. In this paper, we introduce a technique which allows the user to choose a few scene as important according to the application. Based on these selections, our algorithm automatically uncovers the remaining important scenes in the video. In this method, each scene as represented by a couple of numerical values. Since the processing is based on these numbers, it is highly efficient. Our experimental results indicate that this approach performs accurately, and is suitable for large video archives.