Towards real-time monocular video-based avatar animation

Abstract

Most of the commercial motion capture techniques, be they marker-based or marker-free, are either expensive or require complex laboratory setup. These constraints limit their use for home entertainment or interactive gaming oriented applications. In this paper we present a real-time system for color marker-based human motion capture in a monocular image space for avatar animation. The proposed system requires a single webcam and cheap color markers for its operation resulting in low cost and user friendliness. Real-time efficiency is achieved by tracking through variable-sized window search based on a process noise-adaptive Kalman filter. The depth information is recovered with the aid of human anthropometric information in a perspective projection model. Finally experimental results of the system are presented as tested with a number of video sequences. © 2008 Springer Berlin Heidelberg.