As well known,
along the past decades the idea of projecting 3D movies has died
and resurrected
periodically. This time, this idea seems to be here for a long
stay. The proof is the
current competition in the development of 3D TVs or 3D monitors.
Note, however, that it is not very probable that
people accept the need of using special goggles every day at the
home living room. For this reason, we believe
that the next generation of 3D TVs are going to be auto-stereoscopic.
Within the group of auto-stereoscopic displays, the most
glamorous and fascinating are holographic displays. However, at
present there are still some serious problems with this
technique: The need for coherent illumination for pickup and display
stages, and the inherent speckle problem;
The need for reversible holographic materials; and finally the huge amount of data associated to holography.
On the other extreme of solutions we find the auto-stereoscopic
displays based on the codification of two views of the 3D scene
in two entangled series of strips. This technique is optimum in
the sense that the required amount of data is only twice larger
than in classical 2D monitors. But it has two drawbacks:
one is that it provides the observer
with only one perspective, which can be viewed from only one
position. The other is the convergence-accommodation
conflict, which
produces vision fatigue and sometimes headache
An attractive alternative to these techniques is the so-called
Integral Photography (or Integral Imaging), which was initially
proposed by Lippmann in 1908, and which resurrected about two
decades ago due to the fast development of electronic matrix
sensors and displays. The Lippmann idea was that one can record
in a 2D matrix sensor many elemental images of a 3D scene, so
that any elemental image stores the information of a different
perspective of the object. When this info is projected onto
a matrix display (like LCD or OLED) placed in front of an array
of micro lenses, any pixel of the display generates a conical
ray-bundle. And it is, precisely, the intersection of
ray-bundles which produces the local concentration of
light-density that permits the reconstruction of the scene. This
reconstructed scene is perceived as 3D by the observer, whatever
his/her position.
One of
the problems of
integral imaging
systems when working in their standard configuration is that
they provide images that are pseudoscopic; that is, are reversed
in depth. To solve this problem we reported a technique for
formation of real, undistorted, orthoscopic integral images by
direct pickup. The technique is based on a Smart Mapping of
PIxels (SPIM)
of the elemental-images set.
More recently we have
exploited the latent possibilities of the SPIM and have reported
a much more general algorithm. The Smart Pseudoscopic to
Orthoscopic Conversion (SPOC)
allows the
pseudoscopic to orthoscopic transformation with full control
over the display parameters such us the pitch, the focal length
and size of the microlenses, the depth position and size of the
MLA, the depth position and size of the reconstructed images,
and even the geometry of the MLA
Some of the
main challenges faced by our group in order to improve the
performance of InI technique are: to avoid the faceting, to
improve the lateral resolution, to enhance the depth of field by
prevention of the facet braiding and to enhance the viewing
angle of InI monitors.
The first problem (the
faceting effect) results from the from vigneting
occurred when the observer looks through the microlenses. What
happens is that the observer sees only a small portion of the
reconstructed scene through any microlens. The global effect is
that the observer sees a kind a puzzle in which any piece is the
facet seen through the corresponding microlens. Typical faceting
effect occurs when the filling factor of the microlenses is
smaller than one. In such case, any gap between microlenses
inherently results in empty space between facets in the retinal
image. We can then conclude that the use of amplitude modulation
masks for changing the transmission properties of the
microlenses, is highly non recommendable in the display stage on
any InI implementation.
The
lateral resolution of captured elemental images is mainly determined by
sensor
constraints and also by
diffraction
effects. To improve the lateral resolution of InI system in
the capture stage, we have suggested the application of a
hybrid
technique that is based on the use of binary amplitude
modulation during the capture process and Wiener filtering of the captured
elemental images by computer processing. To illustrate the method, we have
calculated the elemental images of four spoke targets for the pickup architecture shown
down in the left In the center we show the
reconstructed images obtained with the proposed hybrid method and in the
right the images reconstructed with the
conventional setup.
The main limitation in depth of field (DOF)
in an InI device appears in the display stage due to the conflict between
the imaging capacity of the microlenses and the ray-bundles intersection
nature of InI reconstruction. As we see in the animation down, a point
object within the reference plane (point
A)
produces a set of sharp elemental images on the sensor. Other points (B
and
C
for example) produce blurred elemental images. In the reconstruction stage,
the elemental images are reproduced on a matrix display placed in front of
the microlenses. The conical ray-bundles generated by any pixel intersect
constructively at the object positions giving rise to the 3D reconstruction.
However, since the matrix display is still conjugate with the reference
plane, all the conical bundles focus in such plane. When the observer looks
through the microlenses, its focus is, naturally, adjusted to the reference
plane, and therefore what it sees is a matrix of bright dots.
This effect is known as the
facet
braiding phenomenon, and imposes the most important limitation to the
reconstruction of long depth 3D scenes. As shown in the down
(a) animation, the detrimental consequences of the facet braiding are more
apparent when observing continuous 3D scenes. In this experiment the
reference plane was set at the optometrist-doll position. The braiding is
really apparent in the windows and chimeny of the house. We have
recently shown that
braiding-free integral imaging is
possible. To obtain this, is is necessary to adjust the display setup so
that the matrix sensor is conjugate with the infinite. Proceeding in this
form, some lateral resolution is sacrificed, but the quality of the
reconstructed image is homogeneous (b).
(a)
(b)
Other problem of InI realizations in the
overlapping between elemental images when working with large scenes. To
avoid this it is conventionally proposed the use of physical barriers
between the elemental cells. An alternative solution, recently reported by
our group, is the implementation of the barriers by purely optical means. In
particular we suggested the use of
Telecentric RElay Systems (TRES).
Due to the
telecentricity of the relay, the aperture stop is back-projected virtually
onto the front focal plane of any microlens. Then, only rays passing through
the projected micro pupils; therefore, emerging the microlens parallel to
its optical axis, will reach the sensor.
The
TRES concept can be used for multiple
applications. One application is the
enhancement of
the viewing angle by of Integral Imaging monitors by use of
Multiple Axis Telecentric
RElay
System (MATRES).
Since the
TRES can be used for projection of any amplitude transmittance
modulation for the microlenses, and therefore for parallel
apodization of the lenses, one of the proposals is to project
the transmittance of a lens of variable optical power to produce
the
micro-zoom array. This
permits the fast, electronic focusing of objects located at
different distances.
For experimental verification, we inserted the a VARIOPTIC
ARTIC-314 liquid lens at the aperture stop of a camera lens, and
combined with a large-diameter lens in telecentric manner. With
this setup we obtained a series of elemental-image sets in which
the in-focus plane was tuned.
|