The Oxford parallel architecture for 3D vision

A hybrid architecture for machine vision is described. The primary components of the architecture are a Datacube pipelined image processor, a configurable network of 32 T800 transputers, a Sun-4 workstation and a special-purpose interface connecting the Datacube to the transputer network. The implementation of a 3D structure-from-motion vision algorithm (Droid) on this architecture is described. This algorithm reconstructs 3D structure by analysing image sequences obtained from a moving camera. The Datacube handles the image digitisation, storage and display; the transputer network performs the feature extraction (corner points) in parallel and the Sun-4 computes the 3D-isation. In this application, which was demonstrated live during a recent ESPRIT conference in Brussels, the architecture delivers a performance of greater than 1 frame per second-17 times the performance of a Sun-4 alone.<
>