Abstract
In April 2004 we built a 198-view multiview volumetric 3D display by modifying the projection surface and rendering software of a swept-screen Volumetric 3D Display (the Perspecta Volumetric Display shown above). An unmodified Perspecta displays a volumetric 3D image by deconstructing a 3D model, such as a computer tomography (CT) scan or molecular model, into a series of 768 pixel x 768 pixel slices. The projection surface is an omnidirectional diffuser with nearly equal transmission and reflection coefficients. Its 25 cm diameter disk-shaped active area is oriented with its normal parallel to the floor and rotates at 900 rpm, centered on the axis of rotation. A 6000 frame/sec DMD projector (Texas Instruments, Inc., Plano, Texas) is used to project 396 slices per rotation that are perceived as a superimposed volumetric 3D image. To ensure proper focus regardless of screen angle, the screen is illuminated by a series of fold mirrors that rotate with the screen.
To create an occlusion-capable 3D display, we replaced the projection surface of a Perspecta display with a diffuser (Physical Optics Corporation, Torrance, California) that has preferentially vertical diffusion and limited horizontal diffusion advertised to be 60° x 0.1° (vertical x horizontal). The vertical diffusion acts to broaden the exit pupil’s vertical extent, providing a wide vertical viewing zone. The restricted horizontal diffusion permits light to exit the display surface with only minimally broadening, allowing the exit pupil to be swept sequentially in the radial direction as the relay optics rotate. By synchronizing projected images with the exit pupil position, we can present a 3D image whose appearance varies with viewing direction over the full 360° in the horizontal direction. To render imagery for viewing on the 3D display, we render a sequence of images using a virtual camera array placed along a circular track surrounding the scene. In 2007, Researchers at USC built a similar 3D display based on this concept, introducing improved rendering algorithms.
Publications
"Occlusion Capable Volumetric Three-Dimensional Display"
O. Cossairt, J. Napoli, S.L. Hill, R.K. Dorval, G.E. Favalora
Applied Optics, Mar. 2007, vol. 46, pp 1244-1250
[PDF]
"Radial multiview three-dimensional displays"
O. Cossairt, J. Napoli
US Patent 7,277,226; Oct. 2007
Images
Here we show images taken from different viewpoints for several 3D scenes loaded onto the 3D the Occlusion-Capable Volumetric 3D display. Each 3D scene was generated by rendering a computer model from a sequence of perspectives, then loading the rendered images onto the 3D display. The images shown are not renderings, they are photographs of the 3D model loaded onto the 3D display. As the viewer changes perspective, the appearance of the scene changes to give the impression of being able to look around objects.
Snowman Scene
A frontal view of a CG snowman.
A side view of the snowman. Notice that the left arm and side of the face is occluded by the body.
Perspecta Logo
A frontal view of a 3D logo. The text reads “Perspecta by Actuality”. Notice the lower text “by Actuality” is mostly occluded by the torus.
A side view of the Perspecta Logo. Notice that the letters “al” are no longer occluded while the letters “by” are now occluded.
Planets Scene
A frontal view of a 3D model of four planets.
A side view of the 3D planets scene. Notice that the furthest planet is now occluded by the planet closer to the camera.
