Rendering, Adaptation and 3D Displays - 3DTV: Processing and Transmission of 3D Video Signals - page 129

On the other hand, there are several advantages of exploiting dense
disparity/depth map information. One of them is that it is possible to
synthesize virtual target views at a further distance from the source (real)
camera views with sufficient perceptual quality thanks to the disparity
information associated with each pixel. Hence, multi-view rendering at a
wider baseline is possible, resulting in enhanced movement parallax. Second,
the use of decoded dense disparity/depth information on the client terminal
side avoids the time-consuming computation of disparity correspondences
from decoded textures of multiple views in order to perform rendering. In
the rest of this chapter, the depth-based view interpolation is explained.
Several different 3D video use cases can be listed depending on the source
multi-view format, delivered service and the 3D display technology used,
though two scenarios will be considered in this chapter. In the first scenario,
the user is delivered a stereoscopic video (i.e. the left view video and the
right view video) and the associated per-pixel depth maps of the delivered
stereoscopic video. The user is free to adjust the depth on the stereoscopic
display, so that the disparity between the viewed left and the right pairs
can be changed. The left and the right views are used in combination
with the depth maps to create an interpolated view in between, where
the positioning of this interpolated view determines the depth level in the
watched stereoscopic content.
In the second scenario, the user is assumed to have a multi-view display
that displays multiple viewing angles simultaneously. More information on
the 3D display technologies is provided later in this chapter. In the second
scenario, the user is delivered two more camera streams in addition to the
stereoscopic video-plus-depth pair. These cameras are located on either side
of the stereoscopic camera pair with varying baseline distance from the
stereoscopic camera pair. It is assumed that all cameras are vertically aligned
and that the disparity is only horizontal. Figure 5.2 depicts the camera
configuration.
The stereoscopic baseline is adjustable in both scenarios. Typical bound-
aries are 3 cm (lowest) and 7 cm (highest), where the adjustment is typically
done with the consideration of stereoscopic vision cues. The inter-axial

Stereoscopic
camera pair

Leftmost camera Rightmost camera

L R

Stereoscopic
baseline

Baseline 1 Baseline 2

Figure 5.2 Stereoscopic 3D video rendering input and output