Block Structures and Parallelism Features in HEVC - High Efficiency Video Coding (HEVC) - page 72

Tabl e 3. 7 Coding efficiency improvement for increasing the set of PU and TU sizes as has been
done in the development of video coding standards. The shown average bit-rates are measured
relative to a configuration corresponding to H.262 j MPEG-2 Video (16 16 PUs and 8 8 TUs

only)
Supported PU and TU sizes

Entertainment applications Interactive applications

H.263 & MPEG-4 Visual 2:8 % 4:3 %

H.264 j MPEG-4 AVC 7:3 % 5:7 %

HEVC 26:6 % 37:6 %

restricted to 16 16 PUs and 8 8 TUs and thus corresponds to the block sizes
supported in H.262 j MPEG-2 AVC. A configuration corresponding to H.263 and
MPEG-4 Visual is obtained by additionally allowing PU sizes of 8 8 luma
samples. For simulating the partitioning features of H.264 j MPEG-4 AVC, we
set the maximum CTU size to 16 16 luma samples, allowed all subdivisions
supported in HEVC, except the asymmetric partitioning modes, and additionally
enabled 4 4 PUs by slightly modifying the syntax. Even though this configuration
differs in some aspects from the actual syntax features of H.264 j MPEG-4 AVC, it
still provides a suitable comparison point. Finally, in the HEVC configuration, the
maximum set of supported block sizes was enabled. The average bit-rate savings
relative to the H.262 j MPEG-2 Video configuration are summarized in Table 3.7 .
Additionally, diagrams with rate-distortion curves for a representative sequence for
both application scenarios are shown in Fig. 3.8 . For the tested HD sequences, the
increase of partitioning modes from H.262 j MPEG-2 Video to H.264 j MPEG-4
AVC yields bit-rate savings of 7.3 and 5.7 % for the entertainment and interactive
scenario, respectively. The experiment indicates that most of the coding efficiency
improvements [ 31 ] that are obtained by H.264 j MPEG-4 AVC for HD material are
related to other coding tools, not the additional partitioning modes. In contrast to
that, a significant part of the coding gain of HEVC relative to H.264 j MPEG-4
AVC can be attributed to supporting increased block sizes for motion-compensated
prediction and transform coding.

3.3 Picture Partitioning for Packetization

and Parallel Processing

As already noted above, both HEVC and H.264 j MPEG-4 AVC follow the
same block-based hybrid video coding paradigm. Conceptually, both video coding
standards also share the two-layered high-level system design consisting of a video
coding layer (VCL) and a network abstraction layer (NAL). The VCL includes all
low-level signal processing, including block partitioning, inter- and intra-picture
prediction, transform coding, entropy coding, and in-loop filtering. The NAL
encapsulates coded data and associated information into NAL units, a logical data
packet format that facilitates video transmission over various transport layers.