Image Processing Reference
In-Depth Information
Fig. 3.1 Stereoscopic image pair and corresponding difference image
environments. Nowadays, autostereoscopic systems that don
t require any eyewear
device are emerging, though not quite disseminate in the consumer market. In this
case, image separation is performed optically using parallax barriers or lenticular
systems [
4
,
5
].
'
3.2.1.2 Multiview Video
When more than two views are conveyed to the receiver, the 3D video format is
simply referred as MVV. In this case, MVV acquisition is typically performed by an
array of synchronized cameras displaced horizontally [
6
].
Using appropriate multiview displays, motion parallax is also supported, besides
stereo parallax, meaning that the user is able to see a different scene perspective
when moving the head along the horizontal plane. This is usually achieved through
multiview autostereoscopic displays with or without head tracking [
4
].
3.2.1.3
3D Holoscopic Video
When a dense number of views is captured simultaneously, MVV can be under-
stood as a light-field or plenoptic function sampling [
7
,
8
]. This means that not only
spatial information about a scene is captured but also angular information, i.e., the
“whole observable” (holoscopic) scene.
The concepts of 3D holoscopic imaging were proposed over a century ago by
Lippmann and referred to as integral photography [
9
]. In 3D holoscopic imaging a
regularly spaced array of small spherical micro-lenses, known as a “fly
s eye” lens
array [
10
], can be used for both acquisition and display of the 3D holoscopic images
to capture and reconstruct the visual scene. At the acquisition side, each micro-lens
works as an individual small low resolution camera conveying a particular perspec-
tive of the 3D scene.
At the display side, a similar optical setup may be used, e.g., a high resolution
LCD panel overlaid with a micro-lens array, to project the recorded 3D holoscopic
image and reconstruct the 3D scene in front of the viewer with both stereo and
'
