Wide-range motion capture from panning multi-view cameras

Abstract

In most motion capture systems by multi-view cameras, fixed cameras have been used. Then, the range of observation, which is a common field of views from a lot of cameras, is narrow in general. Allowing the cameras to be mobile, it expects a wide range of observation but needs to know pose of cameras anytime. We propose a novel calibration technique for a mobile camera, where an image from mobile camera matches with a panoramic background image. Once the cameras are calibrated, fitting a human body model to multi-view images and tracking the body establish a wide-range motion capture system.