Semi-Automatic Placement of Annotations in Video

Report ID: 
Vineet Thanedar, Tobias Hï½llerer
2004-04-01 04:00:00


In this paper, we present a framework for the insertion and semiautomatedplacement of visual annotations into video footage. Visuallyoverlaid annotations are very common in telecasts (e.g., statisticsin sports broadcasts, banner advertisements), where they arestrategically placed by hand, or incorporated into the video streamusing elaborate camera tracking and scene modeling techniques.Such information is not commonly available for raw or unpreparedvideos. We look at the problem of automatically placing annotationswithin the space and time domains of unprepared videos usingcomputer vision and image analysis techniques. We use measuresof intensity and color uniformity, motion information, and asimple model of cluttered screen space to determine regions in thevideo that are of relatively minor importance to the human visualperception. We present a taxonomy of visual annotations, a subsetof whose we realize as automatically placed video overlays, implementedin our prototype system named DAVid (Digital Annotationsfor Video). We conclude with an outlook of potential applicationsof these techniques in interactive television.


PDF icon 2004-11.pdf