Rendering, Adaptation and 3D Displays - 3DTV: Processing and Transmission of 3D Video Signals - page 130

distance that specifies the stereoscopic baseline controls the overall range
of depth, or in other words, the depth volume of the reproduced scene.
In order to create proper stereoscopic video, a variety of rules need to be
considered, one of which is to ensure that the captured scene is fitted within
a Comfortable Viewing Range (CVR) of the viewing context, such as the ratio
of screen width and viewing distance [2].
To facilitate a comfortable long-term stereoscopic viewing experience,
all objects in the scene need to stay within this limited depth space close
to the screen. Baseline 1 and Baseline 2 depicted in Figure 5.2 represent
the distance of the leftmost camera and the rightmost camera from the
stereoscopic camera pair in the centre. The outer cameras may or may
not be equidistant from the centre camera pair and usually the Baseline 1
and Baseline 2 are larger compared to the stereoscopic baseline. The larger
those baseline values, the wider the 3D reconstruction space and increased
horizontal parallax. The view interpolation processing required to address a
multi-view display depends on the display itself, such as the total number
of views that the specific display supports. The distance between each view-
pair on the 3D display can be calculated in two different ways. In the first
assumption, the stereoscopic baseline is regarded as the reference. Then,
depending on the number of views supported by the target 3D display, the
positions of the views to be synthesized within the total camera baseline are
determined. Figure 5.3 shows the respective calculation for a 5-view display
and a 9-view display.
In the second assumption, it is aimed at conserving the complete baseline
of the multi-camera set-up, which is the total of Baseline 1, Baseline 2 and the
stereoscopic baseline depicted in Figure 5.2. The distance between each view-
pair depends on the complete baseline of the multi-camera set-up as well as
the total number of the views that the 3D display supports. Figure 5.4 depicts
the calculation for a 5-view and a 9-view 3D display in the second case.
Virtual views within the total camera baseline are interpolated using two
camera views and the corresponding depth maps, which are first converted
to disparity values. For a target virtual view, one reference is on the left-hand

Stereoscopic
camera pair

Leftmost camera Rightmost camera

LR

5 -View 3 D display

9 -View 3 D display

virtual view

Figure 5.3 View interpolation taking the stereoscopic baseline as the refer-
ence