Video Retrieval Technology Using MPEG-7
Practical application of a video scene retrieval technology that can even cue scenes by combining MPEG-7-compliant metadata with an original high-speed text retrieval technology
Practical Application of Video Retrieval
Ricoh has combined metadata compatible with the MPEG-7 international standard for a multimedia data content description system and an original high-speed text retrieval technology to achieve practical application of a video scene retrieval technology useful in business and educational applications. PowerPoint presentations and lecture videos account for many of the videos used in office work or classrooms. The Ricoh technology first automatically creates metadata from PowerPoint slide data linked to video data. Then, by combining this metadata with a high-speed text retrieval technology, it enables the use of the metadata as search keywords not only for lecturer names and slide titles, but also for words recorded in the slide or notes sections of PowerPoint files. This makes it possible to retrieve desired video scenes using a method similar to that used for ordinary documents and to cue video scenes.
Figure 1: Video scene retrieval using MPEG-7 metadata
MPEG-7 Video Content in Offices and Classrooms
The use of video content in offices and classrooms has rapidly increased in step with IT infrastructure development. As video content increases, retrieval has become enormously important for the effective utilization of these video assets.
Ricoh has participated as a member from the initial stage of development of the MPEG-7 international standard. In 2002 we completed a tool that can efficiently describe MPEG-7 metadata for video, and in 2003 we applied this tool to develop the MPMeister* multimedia Web content automatic generation tool, which we market for use at offices and educational institutions. MPMeister promotes the dissemination of video data by simplifying video handling, previously a bothersome task that required specialized knowledge.
Ricoh's video retrieval technology has been used since April 2008 in the electronic library system of the Digital Library of the Nara Institute of Science and Technology. The university has converted to complete automatic recording of class and lecture videos and automatic Web content preparation, creating a video archive for viewing the content via a network. It is also possible to retrieve this video content at the scene level in the same way as retrieving content from books and to playback those scenes.
Ricoh has participated as a member from the initial stage of development of the MPEG-7 international standard. In 2002 we completed a tool that can efficiently describe MPEG-7 metadata for video, and in 2003 we applied this tool to develop the MPMeister* multimedia Web content automatic generation tool, which we market for use at offices and educational institutions. MPMeister promotes the dissemination of video data by simplifying video handling, previously a bothersome task that required specialized knowledge.
Ricoh's video retrieval technology has been used since April 2008 in the electronic library system of the Digital Library of the Nara Institute of Science and Technology. The university has converted to complete automatic recording of class and lecture videos and automatic Web content preparation, creating a video archive for viewing the content via a network. It is also possible to retrieve this video content at the scene level in the same way as retrieving content from books and to playback those scenes.
Automatic Generation of MPEG-7 Metadata
Although there have been previous examples of manual attachment of metadata to videos and content retrieval, the process entails enormous labor. (It is said that 60 hours is required to prepare an MPEG-7 metadata description of a one-hour video.) Also, although there are systems for attaching simple keywords to videos and retrieving content, for long videos considerable time is required to search for the desired scene. Ricoh has realized bother-free, high-performance retrieval by combining the automatic generation of metadata and a high-speed retrieval system developed over many years.
