Motion reconstruction algorithm using multi-view information
Qinkun Xiao, Dan Luo, Haiyun Wang · 2013
In order to more realistically realize 3D human skeleton animation from image sequences of video, a motion reconstruction algorithm using multi-view information is presented. The method makes full use of the markers' information, which is the key technologies including camera calibration, markers extraction, markers tracking and motion reconstruction. In step of tracking markers, a novel multi-view tracking algorithm is proposed. The algorithm is based on two basic steps, one is binocular stereo vision tracking using Kalman Filtering, and another step is multi-view information fusion. The experiment results prove that the proposed method is effective.