Capture and Processing - 3DTV: Processing and Transmission of 3D Video Signals - page 50

adjustment to create a common convergence point. A post-production
stage is necessary to transform the acquired camera views.

• Vertical displacement between cameras optical axes should be minimized - Ver-

tical displacement (or vertical disparities) between the cameras would
result in serious visual artefacts in the rendered multi-view video on
the screen, because human eyes can only naturally perceive horizontal
parallax. Thus, all the corresponding epipolar lines within all the camera
views should coincide. Note that it is very difficult to ensure the removal
of all such vertical disparities within a multi-camera rig, where a post-
rectification stage is crucial, which will be explained in more detail in
Section 2.3.1.

• The inter-axial distances between the cameras should be adjusted properly -Itis

important that the multi-camera rig is customized depending on the shot
of the 3D scene. It is essential to avoid too much coincidence between
the Field Of View (FOV) of neighbouring cameras. This depends on the
proximity of the multi-camera rig to the shot scene and thus the applied
focal lengths. At the same time, the views of the outer cameras in the multi-
view array should coincide with the outer view of the targeted displays
(i.e. all captured visual information should fit into the FOV of the display).

• It should be possible to extract accurate depth/disparity information from the cap-

tured multi-view video - Accurate and dense disparity/depth information
per each camera makes it possible to synthesize a high-quality intermedi-
ate viewpoint, either within the actual multi-camera baseline, or outside
the baseline of the multi-camera system. Accordingly, it is much easier to
address the display output requirements of multi-view displays and light-
field displays, which usually comprise outputting many more viewpoints
than what is originally delivered to them (i.e. synthesizing M views out
of N reconstructed camera views, where M N ). The depth extraction

process will be explained in more detail in Section 2.3.2.

•

The setting of all cameras should be matched as best as possible in terms
of contrast, brightness, gamut, primaries, etc . - A post-processing stage is
usually necessary to compensate for all such inter-camera differences (e.g.
colour correction that will be described in Section 2.3.1).

Three recent EU-funded projects that are concentrated on an end-to-end
implementation of modern 3DTV chain have also considered several different
multi-view capture architectures. In the following, a basic overview of the
deployed multi-view video capture systems is provided.

2.2.3.1 4-Camera Multi-View Rig of MUSCADE Project

The multi-view acquisition system in MUSCADE is based on four identical
high-definition video cameras. An overview of the deployed acquisition
system is depicted in Figure 2.7. Two of the cameras, the inner pair, form a
regular stereoscopic video pair. This way, it is targeted to directly support