Intelligent autofilming of lectures based on interpretation of the scene environment and evaluating video footage
Masaki Onishi, Masashi Murakami, Kunio Fukunaga · Systems and Computers in Japan · 2006
In this report we propose an intelligent autofilming method as a technique that can be used to film principally blackboard-based lectures automatically, and without human assistance. This method designated the areas to be filmed by interpreting the lecture environment, and, by evaluating the numerous resulting video images, switches to the video images that are ideal for use as lecture images. First, in order for the lecture environment to be interpreted, information pertaining to the lecturer and the matter written on the blackboard is extracted from lecture images that have been shot with a stationary camera, and is used to extrapolate the lecturer's actions. Next, the area to be filmed is determined for each camera based on the lecturer's actions, resulting in footage captured from several camera positions. Last, an assessment is made of each of the numerous clips, based on which the footage that most effectively represents the current lecture environment is selected. We performed an actual experiment in which a lecture was automatically filmed, and confirmed the efficacy of our technique. © 2006 Wiley Periodicals, Inc. Syst Comp Jpn, 37(11): 88–99, 2006; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/scj.10337