An Exploration of Non-tessellated 3D Space Carving for Real-Time 3D Reconstruction of a Person through a Simulated Process

Real-time 3D reconstruction of people promises to combine the winning characteristics of video conferencing and immersive collaborative virtual environments. However, it remains difficult to obtain a reconstruction of a person with recognisable facial features in real-time, as the person moves around. Three highly related issues that require further work are camera placement, reconstruction algorithm and rendering processes. In order to speed up development and concentrate on core principles we have developed a series of software simulators in which all three can be varied without the need to reconfigure a physical system. We introduce these simulators and report evaluation of camera placement, 3D reconstruction algorithms, and texturing. The approach taken is not to create a surface on which to texture but to directly use image data for shape throughout and specific graphics shader functionality to render, light and texture it. The acid test of performance was comparing what our simulated approach could achieve in terms of frame rate and pixel density and comparing this to state of the art commercial “tele-present” video conferencing.