3D-DISPLAYS FOR SPATIAL ILLUSION WITHOUT VISUAL AID
Computer holographic video and volume displays are fascinating new
technologies. Yet the red-green, polarising and shutter glasses have
not been really accepted by the user to date - they are prostheses and
affect both comfort and image quality.
The most important information for depth perception comes from stereo
vision: the eyes transmit two different views of the same scene to the
brain. The same goes in turn for reproduction: each eye has to be
presented separately with a view, by means of mirrors or, more commonly
nowadays, with colour, polarisation or shutter glasses.
But there is another way: "autostereoscopic" displays do without such
spectacles. They use other methods to transfer a left and right image
to the respective eyes. Computer holograms and volume displays go even
further: they draw three dimensional pictures directly into space,
without any stereoscopic tricks.
Among the old but largely unknown ideas is the technique introduced by
Sanyo as "Image-Splitter" at this year's fair (Funkaustellung).
Although Sanyo has registered a new trademark, the technology has been
known since the beginning of the century. A single image contains the
two views of the stereo pair, as vertical stripes which are as narrow
as possible and which alternately show pieces of the left and right
view. A stripe mask in front of the image ensures that the left and
right eye of the viewer see only the stripes dedicated to them, as long
as the viewer keeps his head at a certain distance from the mask.
There is no space for a second viewer.
The three models introduced by Sanyo on the IFA (International
Communications Fair) work with LCDs (4.3 to 10.4 inches) with
left/right stripes that are only one Pixel wide. In front of the
screen is an almost microscopically fine grid, the image splitter.
Dimension Technologies (DTI) has offered a 3D-Display since 1991 that
utilizes the same principle in a modified version. A colour-TFT-LCD
shows the view for the left eye on the pixel columns with even numbers,
and the view for the right eye on the odd number columns. A clever
background lighting of 320 white parallel vertical lines renders a
light absorbing stripe mask superfluous. Special about it is that the
display can be run as a normal 2D-Display (640 columns x 480 rows) or
in 3D-Mode (320 columns).
A model by Terumo (Split-Image-Display) uses two TFT-LCDs to produce
the stereo pair. A half-mirror combines both pictures thus preserving
the full column resolution. Two additional monochrome LCDs provide the
stripe lighting behind the TFT-LCDs. A tracking system measures the
head position and adjusts the viewing zones by moving image display and
stripe display against each other. Instead of blocking the view to one
half of the image pair with a mask the view can also be enhanced
specifically. By means of a dense row of half-cylinder shaped lenses
that directs the two part-images exactly into the corresponding eye.
Such lenticular systems, popular in 3D postcards, have been known since
1908. The method can be refined; if more than two views are used as
image stripes the picture can be viewed from different angles
(parallax-panoramagram). Kodak and the less well known company Art
offer productions of such computer calculated 3D-hardcopies.
Displays based on lenticular systems come from Sanyo and the Henrich
Hertz Institute (HHI), Berlin. Both developments are back-projection
systems. In the Sanyo product two LCD video projectors project the
images for the left and right eye onto a screen (40 inch diagonal).
Stripe lenses are located in front of and behind the screen, a double-
lenticular system. The rear lens system breaks the two projected
images into stripes, the second lens system at the front of the screen
guides each stripe into the corresponding eye of the viewer. To
experience the 3D-effect, the viewer has to keep at a certain distance
with this method (1 metre for the Sanyo product). At this distance
from the screen the distance between the ray bundles is the same as the
distance between the eyes.
But unlike the Image Splitter, the lens system creates several ray
bundles for each spot of the screen; therefore there are several
"stereo windows" and theoretically several viewers can sit in front of
the display (currently actually only two). With the product the
manufacturer is targeting applications in museums, 3D-mini cinemas and
Unlike the Sanyo construction, the HHI display is equipped with a head
tracking system that tracks the head of the viewer. An
electroluminescent flat screen is being developed as well; it also uses
the double lens system and it achieves HDTV resolution.
Together with the company Carl Zeiss, the HHI is currently developing a
3D flat screen for use in medical technology. With the new
development, the viewer supposedly is able to move about thanks to a
movable lens system plate.
Since 1942 Sharp has used a different concept of 3D-displays called
twin-LCD. This system projects the images of two LC-displays
separately into the eyes of the viewer. Particularly interesting in
this set-up is the folded ray course. Using a half-permeable mirror
the system can do with a single light source. If the viewer moves it
is sufficient to move the light source in the opposite direction, and
the rays meet the viewers eyes again. The light source is guided by a
tracking system that not only reacts when the viewer approaches the
screen or moves away from it, but also when he moves sideways. Image
sources for the twin-LCD are either two laser disk players or a
graphics workstation connected to the tracking system. This
workstation calculates, in real time, lateral views of the scene
enabling the viewer to have a dynamic angular view (motion parallax) up
to 40 degrees around the object as long as he does not move too fast
(up to 30 cm/sec).
A plan for the future: with further light sources in the optics several
viewers could sit in front of the twin-LCD.
All these stereoscopic displays do not create a real three dimensional
image but a spatial illusion, since the human spatial vision can be
deceived with two flat pictures. Serious problems can arise from the
conflict between accommodation and convergence. The partial images
have full depth of field with objects appearing focused at any
distance. Gradual loss of the spatial impression, head and eye aches
and even nausea and vertigo could be the results. Also, by moving the
head one does not gain new views, the movement parallax is missing. As
in the Sharp development it can be simulated with a computerized
tracking system. The three particularly important spatial informations
for three dimensional viewing are, - apart from movement parallax-
accommodation, convergence and binocular parallax; they are created
only through actual objects or real 3D-Images.
The fault list of stereoscopic displays is long: fixed viewer position,
only one or two viewers, no 360 degrees viewing. To overcome these
problems 3D-images have to be drawn into the real space. To depict
three dimensional scenes in space, holography is the most popular
technique. The usual method of creating holograms is to split a laser
light bundle in two parts, of which one reaches the film directly
(reference ray) and one is reflected from the object onto the film.
The film itself does not carry a directly recognizable image as
information but only a phase difference, an interference pattern of
bright and dark lines holding the coded spatial information. The 3D
image appears when the hologram is lit appropriately, suspended in
space and seemingly real like an object. Depending on the recording
method a hologram can also store the colours of the photographed
scene. Today, high resolution reproductions in natural colours are
only possible from actually existing objects. To be able to animate
computer models in space like on a screen would be video-holography.
Wouldn't it be possible to simulate the holographic recording process
on computer, track the course of the light wave front and thus create a
synthetic computer-hologram, without laser, film and a real object?
The Spatial Imaging Group of the Massachusetts Institute of Technology
began working on the problem four years ago and their results are
Starting points of the calculations are 3D-objects created with CAD
software. A computer generates a hologram from this data which is
stored in an image memory. With a special real time-holography-display
("holovideo") the three dimensional image can be produced in real
time. The computer-holographic pattern is transferred as acoustic
oscillations into a crystal acousto-optic-modulator (AOM). The
oscillations change the refraction index of the crystal both
geometrically and timewise. A laser traveling through the AOM is
modified according to the holographic interference pattern similar to a
normal hologram on a photographic plate, and a 3D-image is formed in
space. An additional optical deflection system reflects the synthetic
hologram row by row to the viewer, with a rate of 40 pictures per
The biggest problem of this technique is the huge amount of data. An
object of 10x10x10 cm with a 30 degree viewing angle makes a hologram
of 25 Gigabyte; the required data transmission rate at 60 pictures per
second in 8-bit-resolution would be 12 Terabytes. So far workstations
have been occupied for minutes to hours creating the computer generated
holograms (CGH), even with the necessary massive cut backs in
Real time video holography was achieved with a connection machine and
data reduction (such as neglect of the vertical parallax, restriction
to 11 degrees viewing angle, 3 cm object size). New coding techniques
reduce calculating time to a hundredth and enable workstations to
calculate 6-Mbyte-holograms in less than two seconds. Fully coloured
interactive video holography in real time has come within reach.
The ideal three dimensional depiction is an image in space that one can
walk around and view the object from all sides. Image generation in
space (as opposed to one or two planes) could be achieved by stacking
sections of the object image in rapid sequence into space. But for
this the odrawing boardoe has to fill a volume which is made possible
through projection planes that oscillate through space (multiplanar
displays). The idea has been realized in several ways.
Since the mid-sixties oscillating or rotating projection planes have
been used. One of these concepts utilizes a flexible concave mirror.
Made to vibrate by a speaker membrane, the focal length of the mirror
(varifocal mirror) changes. This way the image of a cathode ray tube
is focused on different planes. The apparent image depth is far
greater than the amplitude of the mirror. In 1988, BBN produced such a
device commercially. A disadvantage of the principle are the optical
distortions of the image; a 360 degree view is impossible.
In 1991, researchers from Canterbury University, New Zealand, took up a
patent from 1964 and improve it: a phosphorus coated disk rotates
inside a Cathode Ray Sphere (instead of the known cathode ray tube).
Two electron guns and a conventional deflection unit make the
phosphored screen glow. When the screen rotates, a translucent three
dimensional image is visible. The electron guns are shifted by 120
degrees and positioned below the equator to generate better 360 degree
viewing. The problem is the limited size and the high demands on the
quality of the vacuum.
Apart from other concepts with rotating projection planes (i.e.
spirals), techniques with self glowing planes have been invented. In
1977 a system by MIT that packed 4096 red LEDs on a rotating disk was
patented. The glow pattern, however, could only be changed during
stops. Future developments aim to integrate the electronics for image
generation and storage on the rotting disk. The mechanical strains due
to the fast rotation limit the technique. Therefore, rotating planes
onto which the image points are projected are of interest again.
In the late seventies Prof. R. Hartwig and a group of Heidelberg
researchers suggested a system with a rotating spiral plane (helix).
Such a 3D-Display with a rotating helix was produced in a collaborative
project of a secondary school in Stade and the "Institut fuer
Flugfuehrung" of Braunschweig University. The projection plane rotates
with 1200 rpm in a glass cylinder. The semi-transparent plane is a 360
degree helix plane. Red, blue and green laser rays parallel to the
rotation axis hit the helix and generate a light point on it that is
visible from all sides. A computer generates the image in real time.
The image resolution however is restricted by the limited laser
deflection frequency and the computing speed.
Kodak also uses the Hartwig principle but substituted the laser system
with a video projector, which required the installation of a
complicated, self-rotating lens to solve the focusing problems.
Texas Instruments' "Omni View" development is working on the same
concept since 1988; a prototype of 36 inch diameter was made available
to the Navy in 1992. The projection plane of this display consists of
two helices wound like mirror images. The plane rotates with 600 rpm
and thus fills a cylinder volume. Five laser systems - one for each
space quadrant and an additional one to produce higher resolution (up
to 15000 voxels, volume picture elements) - with three lasers (red,
yellow, green) each, create coloured three dimensional images visible
from all sides. Light rays are synchronized with the rotation of the
disk and are split into small packets of up to 10000 per second. A
scanner lays the laser rays in a grid pattern on the plane in the x and
y directions; the z coordinate results from the synchronisation with
Disadvantages of all such multi-plane displays are the mechanical
strain and the complicated "targeting". Most have dead zones, dark
space areas that can not be targeted, such as the axes of rotating
projection planes. It would be better if the space points could be
targeted directly, without having to use the rotating plane as a bridge
from the Pixel-flatland to the Voxel-space.
Real volume imaging is still a dream, but the dream has a solid basis -
stepwise excitation of fluorescence. Already in 1920 it was observed
that atoms of mercury vapour can be activated to after-glow
(fluorescence) when exposed to two different subsequent light pulses.
For fluorescence atoms have to have three suitable energy levels; they
are activated at the crossing point of two invisible infrared light
rays from the first to the second level and from there to the third.
Returning to their basic level they emit visible light - the crossing
point of the infrared rays glows. Experiments of that kind were
successful with an erbium enriched calcium fluoride crystal, creating
glowing space points that can be targeted. The glow, however, is only
visible in total darkness. So far researchers have not found an
efficient medium for larger displays. There is speculation about
suitable physical processes other than fluorescence; voxels of
photochemically activated molecules could scatter light in colours.
The ability to see three dimensionally depends on a multitude of visual
informations that are gathered by our eyes and their motorics. Some of
the information is gained by one eye alone, other information requires
The laws of geometrical perspective were already known to Euclid.
Renaissance painters were the first to utilize linear perspective and
the depiction of light and shadow at the same time to give the viewer a
perception of depth. The aerial perspective (the obluing of the
horizonoe) followed and, with Leonardo da Vinci, even depth of field:
The (apparent) size of known objects and the mutual position of several
objects also aid in estimating distances: if an object covers another
object it lies in front of that, not behind it. When we move, the
relative position of objects changes. All these depth informations can
be obtained by one eye.
With two eyes stereoscopic vision is possible, enabling a human to
perceive distance differences with an accuracy of 1:1000. This is
based on the two images received by the retinas. They differ due to
the distance between the eyes. From the differences the brain models a
spatial impression. How this exactly works is not fully understood
yet. The magic pictures and optical paradoxes of an M.C. Escher both
challenge and confuse our visual intelligence.
When adjusting to a point at a certain distance both eyes turn until
their viewing lines cross at that point. The muscle tension expended
for accommodation and convergence also provides very important
information. At distances beyond 6 metres these oculo- motoric effects
disappear; the eyes are focused to infinity and are nearly parallel.
If convergence and accommodation contradict each other then nausea,
headaches and the loss of depth perception are the results.
Translation (1996) of original German article
from: Adam Schwarz