Graphics Reference
In-Depth Information
7.5.2
SAO Coding Efficiency and Subjective Quality
Improvement
This section illustrates the subjective and objective performance of the SAO tool.
Tab le
7.5
reports the sequence-wise luma BD-rates and the average luma BD-rates
and run-times for different encoding structures and CTU sizes equal to 64
64 in
luma using the skipping boundary samples algorithm as described in Sect.
7.4.2.5
.
For
BQTerrace
in the LP condition, the SAO coding gain reaches 18.9 %. It is
noted that SAO is particularly effective for Class F sequences, which mostly contain
computer graphics and screen content rather than natural video. One could also
notice that SAO shows higher coding gains in the LP configuration without bi-
directional prediction. Regarding the computational complexity, SAO increases the
average decoding time by less than 2-3 %.
The subjective quality improvements due to reduction of ringing artifacts are
shown in Figs.
7.24
and
7.25
. Figure
7.24
shows an example of the coded computer-
generated sequence SlideEditing. SAO significantly improves visual quality by
suppressing ringing
artifacts
near objects edges. Figure
7.25
shows examples of
natural video sequences
RaceHorses
and
BasketballPass
where the edges of objects
are much cleaner when SAO is enabled. According to viewing tests, SAO improves
subjective quality [
42
].
7.5.3
Combined Effect of In-Loop Filters on Coding Efficiency
Tab le
7.6
demonstrates objective compression efficiency improvements due to both
in-loop filters compared to the configuration where both the deblocking filter and
SAO are turned off. One can see that the compression efficiency improvements
are 2.6-15 % depending on coding configurations. The decoding time increase is
about 10 % and depends on the coding conditions. The encoding complexity mostly
depends on a particular encoder implementation and is not significant in the HM11.0
encoder operating in common test conditions (on the order of 1 % encoding time
increase [
13
]). These numbers indicate that the in-loop filters are an efficient tool in
improving the HEVC compression efficiency.
7.6
Main Differences between HEVC and H.264/AVC
In-Loop Filters
This section summarizes key differences between the HEVC and H.264/AVC in-
loop filters. There is only a deblocking in-loop filter in H.264/AVC, while the HEVC
standard defines two in-loop filters: the deblocking filter and the sample adaptive
offset, SAO.
