Shooting the lecture scene using computer-controlled cameras based on situation understanding and evaluation of video images

We propose a computer-controlled camera work that shoots object scenes to model the professional cameramen´s work and selects the best image among plural video images as a switcher. We apply this system to a shooting of a lecture scene. In the first image, our system estimates a teacher´s action based on the features of a teacher and a blackboard. In the next, each camera is directed to a shooting area based on the teacher´s action, automatically. In the last, this system selects the best image among plural images under the evaluation rule. Moreover, we have tried experiments of shooting lecture scene and have confirmed the effectiveness of our approach.