Videopot: Indexing-Based Desktop Video Search System

We present videopot, a desktop video search system designed for digital video files stored on personal computers. The core metric of this system is based on video-indexing technology and an automatic related-term collection framework. Digital files on local disks are crawled, and images of representative scenes are extracted from video files using video-indexing technology. Related terms, used as target information in the video search, are collected from both local files and Web pages using an automatic related-term collection algorithm. These images and terms are then stored in the metadata database. We describe a unique search result interface that helps users distinguish the target video. In the search results interface, the results of the video search are listed at two levels to let users search videos faster. The effectiveness of images on our interface derived from video-indexing is evaluated through a comparison with a conventional file search application