Studies of Automatic Video Generation from Real World

Yoshinari KAMEDAMichihiko MINOHKatsuo IKEDA
Center for Information and Multimedia Studies, Kyoto University
Graduate School of Informatics, Kyoto University
Yoshidahonmachi, Sakyoku, Kyoto, Japan, 606-8501, Japan
E-mail: {kameda, minoh}@media.kyoto-u.ac.jp, [email protected]
URL: http://www.imel1.kuis.kyoto-u.ac.jp

Abstract:

We propose two methods that enables us to make videos automatically from daily scenes. One is for shooting human activities held in a fixed place like lectures, plays, and so on, and the other is for reviewing video taken at a place where people go by.
In the former method, multiple active cameras are needed to shoot multiple objects in the scene. Therefore, a new cinematography is required that can cope with many active cameras simultaneously and can select the best video according to the situation of the scene. We have been developing automatic camera system which estimates the situation of the scene based on image understanding technique and controls the cameras so as to obtain the best shots of the activity.

With the latter method, videos presented to them are to be skimmed so that video viewers can save their time. We have been working on this issue and have done experiments for years and discuss the way to use stored video clips.

Yoshinari Kameda
Tue Oct 5 01:12:48 JST 1999