Capture and Processing - 3DTV: Processing and Transmission of 3D Video Signals - page 55

Pixel groups

Perpendicular lenticular sheet
attached to the surface of the
LCD display

Reconstructed 2D viewpoints

Figure 2.11 Toshiba's multi-parallax glasses-free 3DTV display system deploy-
ing integral imaging principles

Toshiba delivers nine parallax images that are refracted into nine different
spots. In order to create these views, each pixel of the display panel is
clustered into a group of nine, which then can display the same detail of the
picture from nine slightly different viewing angles. These pixel clusters are
covered with an orthogonal lens array to refract the generated pictures at the
correct angle pattern. Figure 2.11 depicts this example.
Although integral imaging capture systems are convenient from the point
of view of the capture set-up adjustment framework, it is not possible to
capture a scene with a larger baseline than a carefully designed multi-view
camera rig (i.e. the viewpoint navigation capability is very limited to the
vicinity of the central viewpoint of the capturing camera).

2.3 3D Video Processing

The video streams captured by a rig of cameras should be processed prior
to encoding and transmission, in order to balance the optical differences
between different cameras, align them by transforming their coordinate sys-
tems and adapt the video streams to the external capturing conditions, such
as colour distribution and exposure parameters. Most 3D video process-
ing frameworks start with the calibration of the individual cameras within
the multi-camera capture rig. In other words, each camera is calibrated
separately first.
Camera calibration refers to the process that determines all the relevant
camera intrinsic parameters that describe how an image is captured. Thus, it
is possible to deduce the real-world coordinates of an object from its image