With the rapid development of Internettechnology and the prevalence of easy-use video capturing devices, the amount of videos is growing explosivelyon the web. Meanwhile digital processing tools make itquite easy to copy, reformat, modify and republish "new"video files, resulting in a large volume of duplicate content.Web video duplicate detection and measurement can facilitate effective video search and intellectual property protection. In this paper we mainly investigate this problem. Wefirst explain the concept of video vocabulary and employa heuristic searching method to build it. Based on videovocabulary, we propose two kinds of video signatures thatare robust to changing compression formats, compressionratios, frame sizes and frame rates and then develop thecorresponding signature comparison schemes. The experiments demonstrate that our approaches are both effcientand effective for web video duplicate detection, even forreal-time applications.