IEEE Consumer Electronics Magazine - September/October 2022 - 18

Overview of VVC for Consumer Electronic Products
Table 1. Per-tool VVC performance.
by the JVET document,23 which demonstrates that
tools' performancewas not optimized for the JVET
test sequences. LMCS presents lower gains in this
articlethaninChen et al.'s work,23 which tends to
show that this tool requires very accurate and content-dependent
tuning. Gains fromSAOaresmall in
terms of objective metrics, but this tool was considered
by JVET as particularly relevant for subjective
quality. Also, its implementation is extremely
low cost, which makes this tool very useful when
ALF is disabled.
In Figure 4, the tools listed in Table 1 are
Notes: The BD-rate performance are reported by disabled
tool in terms of PSNR, VMAF, and SSIM. The relative
encoding time (EncT) and decoding time (DecT) variations
are reported with respect to the anchor.
y Intra-tools, transform tools, in-loop filter tools, ?
MV
coding, sub-block motion comp, motion compensation,
prediction enhancement.
variation indicates a bit-rate increase when disabling
the tool. Encoding and decoding runtime
variations are also reported. These latest figures
are provided as indicative data, but must be considered
with a lot of care, since the VTM decoder
implementation is far from a real product implementation
and from an efficient implementation.
Detailed results are reported in Table 1. Tools are
grouped per category (y intra-tools, transform
tools, in-loop filter tools,
?
MV coding, subblockmotion
compensation, prediction blending,
prediction enhancement). When all these tools
are disabled, BD-rate variations are 43.18% for
PSNRYUV, 30.22% for VMAF, and 27.67% for MSSSIM,
with encoding and decoding runtime variations
of 19% and 49% compared to the setting with
all tools enabled (anchor). This demonstrates the
substantial coding performance brought by the
new VVC tools. It is also observed that, for most of
the evaluated tools, the reported results are very
consistent with the per-tool evaluation reported
18
grouped per category, and relative contribution
of each category to the total PSNRYUV BD-rate
coding gain brought by all the tools is depicted.
Similarly, the relative runtime ratios are depicted
in Figure 4(b) and (c) for encoder and decoder,
respectively. All tools' categories turn out to have
an important contribution to the overall coding
gain. Most computing demanding parts in decoder
are loop filtering and inter-coding (motion compensation,
sub-block MC, and prediction enhancement).
Intra- and transform categories have
negligible impacts on the VTM decoding time. At
VTM encoding side, the most demanding part is
the inter-part, which represents a 2/3 of the encoding
time increase. Another substantial part of the
encoding runtime is consumed by the partitioning,
which is not assessed and reported in this article.
The decoding runtime impact of these tools
remains limited in this particular RA coding
configuration.24
Real-Time Implementations
As of now, VVC benefits from both industrial
and open-source implementations, contributing
to the emergence of an end-to-end value chain.
For example, manufacturers, universities, and
research institutes have announced the availability
of fast VVC implementations.25 Among existing
solutions, the INSA Rennes real-time OpenVVC
decoder, the Fraunhofer Heinrich Hertz Institute
VVdeC decoder,26 VVenC encoder,27
and the
ATEME TitanLive encoding platform have been
developed during the last months.
REAL-TIME VVC DECODING WITH OPENVVC
OpenVVC is the world-first real-time software
VVC decoder developed from scratch in C
IEEE Consumer Electronics Magazine
IEEE Consumer Electronics Magazine - September/October 2022

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - September/October 2022
