Compression - 3DTV: Processing and Transmission of 3D Video Signals - page 74

6-tap filter. In addition to the increase in taps, HEVC always uses 14-bit
accuracy in motion compensation, irrespective of the bit-depth of the video
source material.

3.3 3D Video Coding

L-R stereoscopic video, the simplest form of 3D video, requires more storage
capacity and higher bandwidth for transmission compared to 2D video.
Therefore, 3D video coding is crucial to make the immersive video applica-
tions available for the mass consumer market in the near future. The coding
approaches for 3D video may be diverse depending on the representation of
3D video. 3D video coding approaches aim to exploit inter-view statistical
dependencies in addition to the conventional 2D video coding approach,
which removes the redundancies in the temporal and spatial dimensions.
The prediction of views utilizing the neighbouring views and the images
from the same image sequence are shown in Figure 3.5. The efficiencies of
the prediction methods shown vary depending on the frame rate, the inter-
camera baseline distances and the complexity of the content (e.g. spatial and
temporal characteristics).

3.3.1 StereoscopicVideoCoding
Many stereoscopic video coding algorithms have been proposed and tested
till now. A common approach has been to encode one of the viewpoints
(either left-eye or right-eye view) conventionally using conventional video
coding tools (e.g. AVC), and the other view using disparity compensated
prediction. This is a fundamental way to exploit the 3D scene geometry
in improving the rate-distortion efficiency of the stereoscopic video codec.
Several disparity-compensated prediction-based stereoscopic video coding
works exist in the literature, such as in [11] and [12]. It is important to

S n − 1 S n S n + 1

First - order neighbours :

T n − 1 T

S/T S/T

- T ( temporal )

- S ( inter-view )
- S/T ( combined )

P

T n

S S

T n + 1

S/T T S/T

Figure 3.5 Statistical dependencies in a three-dimensional video