Compression - 3DTV: Processing and Transmission of 3D Video Signals - page 84

According to the test results, it has been found that MVC that is based on
AVC beats the platelet-based multi-view depth codec, as well as the AVC
intra-coder, with a large margin in terms of rate-distortion performance. It
is, however, the case, if the distortion of depth map reconstruction is taken
into account during the in-loop rate distortion optimized mode selection
process. On the other hand, when the reconstruction quality of arbitrarily
synthesized virtual viewpoints is taken into account, the performance of
the platelet-based coder beats that of the AVC intra-coder and comes close
to the performance of the AVC-based MVC at various depth map coding
rates. It is necessary to preserve the edge information in depth maps, in
order to facilitate better rendering of views, where edges in depth maps
correspond to sharp depth discontinuities. Taking into account subjective
quality assessment, the authors come to the conclusion that the platelet-based
multi-view depth map coder preserves the sharp depth discontinuities more
than MVC does, and hence introduces less coding-related structural artefacts
in the synthesized viewpoints.
In [40], the authors compare two different mode decision techniques in
depth map encoding using the AVC standard, where the distortion measure
for depth maps was selected as the distortion in the resultant synthe-
sized views using the corresponding compressed depth maps. Both of the
compared depth-block coding mode selection approaches relied on the men-
tioned depth/disparity quantization error metric, as shown in Figure 3.12,
which takes into consideration the amount of error introduced in the synthe-
sized view. Subsequently it has been discovered that taking the distortion
measure of a compressed depth map as the resulting structural error in the

Distorted
synthesized
video

Compressed
depth

Depth image-
based rendering

Color (Original)

Sum of squared error (SSE)
to be used as the distortion measure

AVC
ENCODER

Subtract

Distorted
synthesized
video

Depth image-
based rendering

Original depth

Color (Original)

Figure 3.12 Depth distortion measure as the SSE between the original and
distorted synthesized view (as described in [40])