3D or not 3D

Toby Howard

This article first appeared in Personal Computer World magazine, Frbruary 2001.

If there ever was a sci-fi dream, it's to conjure up images floating in space – true 3D images that you can view from any angle. Who can forget that classic Star Wars moment when Princess Leia, projected as a shimmery hologram by R2D2, asks for Obi-Wan's help? Now the dream may be getting closer to reality, with Massachussets start-up Actuality Systems Inc. sayingd ot has a revolutionary new product shipping "soon" – a display capable of creating 3D images that hover in space.

Despite much research, especially at MIT during the 1990s, the kind of projected video holography used by R2D2 isn't yet a reality, but there are other technologies. Most, such as the autostereoscopic displays developed by Dimension Technologies and Sharp, are based on the principle of presenting each eye with a slightly different view of the same image, and letting the brain perform its miracle of 3D perception. Clever optics and head-tracking mean there's no need to wear any special glasses to keep the two images separate for each eye.

But such displays aren't truly 3D – the 3D-ness of their image exists only inside our heads. Making "volumetric" displays that really do display 3D images in 3D space is a much tougher problem. In principle, a volumetric display is an obvious extension to the conventional 2D display. Just as a conventional display lights up pixels in a 2D grid, a volumetric display renders images using a 3D grid of "voxels". And like a pixel, each voxel can be assigned a colour and intensity, The really hard part is engineering the voxel grid.

Recently, engineers Barry Blundell and Adam Schwarz published the first-ever book about volumetric displays, and it's a fascinating read ("Volumetric Three-Dimensional Display Systems", Wiley, ISBN 0-471-23928-3). Incredibly, the idea dates back to 1912, when two French inventors, Emile Luzy and Charles Dupuis, obtained a patent for their proposed system which created voxels at points in a solution of selenium chloride, illuminated by intersecting beams of infra-red light.

Today, the focus is on using "swept volume" methods, which use virtual voxels. The voxels don't really exist in space – we only see them at all because of our persistence of vision, which lets us "see" an image for about 1/30th of a second after the actual image is removed from view. The idea is that a rotating 2D screen displays a series of 2D slices of a 3D image, and the each image trace left by a pixel becomes the virtual voxel.

It's on this principle that the Actuality Systems display seems to work. According to their Website, their display takes the 3D image data as a set of 200 or more 2D slices, each of which is a disk of diameter 768 pixels. A rotating mirror reflects each slice in turn onto the moving display screen, which, with the help of our persistence of vision, sweeps out a trail of images which last long enough to give the impression of a real 3D image. The system will have 6Gb of display memory, to cope with 90 million virtual voxels, although there'll only be 8 colours. And there'll be an OpenGL programming interface. There's no mention yet of cost, but it isn't likely to come cheap. Company founder Gregg Favalora told me: "We expect to begin low-volume production sometime in 2001".

If Actuality Systems comes up with the goods, and I hope they do, the next step will be to integrate true 3D imagery and real-time interaction using force-feedback. The glass barrier that separates us from the computer model will disappear forever. One day soon, something like R2D2's hologrammic projections may become commonplace, although the images we'll see are more likely to be drug molecules and CAD designs, than the pleas of an inter-galactic Princess.

Toby Howard teaches at the University of Manchester.