Graphics Reference
In-Depth Information
Fig. 7.24
Example of test sequence
SliceEditing
in LP condition, POC D 100, QP D 32: (
a
)SAO
is disabled, (
b
) SAO is enabled
Fig. 7.25
Subjective quality comparison of
RaceHorses
test sequence, POC
D
20, QP
D
32, LP
condition: (
a
) SAO is disabled, (
b
) SAO is enabled, (
c
) original (uncoded) sequence
The computational complexity of the HEVC deblocking is lower than that of
the H.264/AVC. Reduction of the HEVC deblocking complexity is achieved by
restricting the filtering to the 8
8 sample grid in contrast to the 4
4samplegridin
the H.264/AVC deblocking. Additional complexity reduction in HEVC is achieved
by making the sample-based filtering decisions based on a subset of lines across the
block boundary in contrast to the line-based decisions in H.264/AVC deblocking.
Moreover, the HEVC deblocking of chroma components is only applied to the intra-
predicted block boundaries. The HEVC deblocking is also more suitable for parallel
implementation than the H.264/AVC deblocking since each 8
8 sample block in
HEVC can be deblocked independently of other 8
8 blocks and the order of the
vertical and horizontal filtering operations in HEVC deblocking is always the same.
The processing order for the horizontal and vertical block boundaries is therefore
different in HEVC and H.264/AVC. When applying the HEVC deblocking on the
CU basis, right after the CU reconstruction, filtering of four right-most samples of
horizontal block boundaries in a CU should be delayed until the next CU to the right
is reconstructed and the vertical boundary between the CUs is filtered.
HEVC and H.264/AVC deblocking filters are also different in terms of criteria
that evaluate the signal (reconstructed sample values) at the sides of a block
boundary to decide whether the deblocking is applied to this block boundary. In
H.264/AVC, the deblocking is typically applied to the block boundary when the
