Introduction

The driving goal of our research group is to render 3D scenes using images, rather than polygons, as the source data. This requires full-color images augmented with range data on a per-pixel basis. Each of these images is used as an input primitive to a rendering technique called image-based rendering [2].

In this paper, we review the technology we have developed for capturing depth of an environment. The hardware consists of a time-of-flight laser rangefinder, a line-sweeping mirror, and a panning unit all attached to a conventional PC. Custom software is required to control this hardware and is described. After a brief description of calibration and error compensation, we present several examples, showing both the quality of the range readings and the problems encountered in real-world environments.