Abstract
Content of a movie is produced in two different types of design environments. The first is the design environment of shooting where a camera is used to capture what is happening at a particular place and time. The second is the design environment of editing where the rushes are interpreted relative to a movie maker's intent. Annotation of the video stream allows the movie maker to make decisions based on specific content of video and in the best case enables a machine to help in that process.

Stratification is a context-based layered annotation method which treats descriptions of video content as objects. Stratification offers a graphical representation of the content of a video stream and enables movie makers to quickly query and view descriptions for any chunk of video. Stratification supports the development of complementary or even contradictory descriptions which result when different researchers access video source material which is made available on a common workstation or over a network.

Lecture Notes in Computer Science, No. 712, November 1992, pp. 250 - 261.
http://mf.media.mit.edu/pubs/conference/StratificationSystem.pdf