A GPU based, projective multi-texturing approach to reconstructing the 3D human form for application in tele-presence

This paper reports a GPU based, projective texturing approach to reconstructing the human form, from multiple images, at a quality and frame rate close to high end video conferencing. The ultimate aim is to support spatially grounded, non-verbal communication through a video based medium. This will, we hope, enable us to balance image quality and update rate to deliver highly realistic and dynamic 3D human representations that offer the visual quality of high end video conferencing with the spatial and temporal characteristics of immersive virtual environments. The output of this will enhance communication by enabling a remote actor to be realistically projected into another person's local space, projected into an extension of the local space, or projected into a shared virtual space. This extends previous work by incorporating texture into the reconstructed form and evaluating the optimized process within our established simulation system.