Information Technology Reference
In-Depth Information
2.4 Discussion of Stereoscopic Systems
Stereoscopic techniques are definitely the simplest and cheapest, thus the most
widespread methods to generate 3D vision. On the other hand, they come with
several drawbacks. A stereo image with glasses provides correct 3D images only
from a single point of view. Observing the same image from other locations re-
sults in distorted views, which is most visible while moving in front of the screen,
when the image ”follows” the viewer. Although this limitation can be overcome
by tracking the position / orientation / gaze of the user and updating images in re-
sponse to movements [21], some latency will inherently be introduced [22], sig-
nificantly compromising immersiveness and limiting the correct view to a single
(tracked) user. This and other missing 3D cues result in effects like discomfort,
sea sickness, nausea and headache which make them inconvenient for long-term
use according to some users [23].
One possible explanation comes from neuroscientists' research in the field of
human perception of 3D. They found that showing each eye its relevant image is
not enough for the brain to understand the 3D space [24]. For getting the 3D pic-
ture of the environment, humans rely on two main visual cues: the slightly differ-
ent image seen by each eye and the way the shape of an object changes as it
moves. A brain area, the anterior intraparietal cortex (AIP), integrates this infor-
mation [25]. With a stereoscopic display the image becomes 3D, but as soon the
brain thinks that it does see a 3D image, it starts working like in a normal 3D
world, employing micro head movements to repeatedly and unconsciously check
the 3D model built in our brain. When an image on a stereo display is checked and
the real 3D world mismatches the 3D image, the trick is revealed. Presumably the
AIP cortex never got used to experience such 3D cue mismatch during its evolu-
tion and this produces glitches which result in unwanted effects.
2.5 Stereoscopic 3D Uncompressed Image Formats
Stereoscopic displays need two images as input (left eye and right eye image),
which seems to be simple, yet various formats exist. The most straightforward so-
lution is having two different images making up a 3D frame (see Fig. 5.), but this
requires double bandwidth compared to the 2D case.
Another common approach uses an image and a corresponding depth image of-
ten called 2D + Depth (see Fig. 6.), which may consume less bandwidth depend-
ing on the bit depth of the depth map, but needs metadata to map depth informa-
tion to the 3D context, and still consumes more than a 2D image.
The 2D + Delta format stores the left (or right) video stream intact, and adds
the stereo disparity or delta image that is used to reconstruct the other view. The
advantage is that compressed Delta information can be embedded into an MPEG
stream in a way that does not affect 2D players, but provides stereoscopic infor-
mation to compatible 3D decoders [26].
To make the transition from 2D to 3D easier, broadcasters and manufacturers
preferred stereoscopic image formats that can be fit into a 2D frame compatible
