Abstract
Video taken from a single monocular camera is the most common means of recording human motion. In this article, we present a practical, semiautomatic method for synthesizing a human motion that is guided by such video. After preprocessing an input video, we select a precaptured motion clip called a reference motion from a motion library. We then compute the sequence of body configurations of a virtual character by deforming this motion, according to spacetime constraints derived from a sequence of 2D features in the input video. Experimental results show that our method can synthesize highly dynamic motions, such as kicking and header motions of soccer players. We also showed the potential of our scheme as a new paradigm for motion capture, that is, capturing motions from videos taken with a monocular camera, even outside a motion capture studio.
Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
