Video‐based Approach to Human Animation

Abstract

A method based on computer vision technologies is presented to determine the 3‐D spatial locations of joints or feature points of a human body from human motion video. The proposed method first applies the geometric projection theory to obtain a set of feasible postures in some key frames according to predefined 2D video features and 3D‐model features correspondence. Next it makes use of the available skeleton controlled human model to get a feasible posture for each key frame. The method is applied to a series of video images to animate artificial 3D human models.