Graphics Reference
In-Depth Information
In case of a CTU-based processing, four lines of samples for the luma component
and two lines for the chroma components should be kept in a line buffer for in a
line buffer both the HEVC and H.264/AVC deblocking. The fourth line of samples
would also be modified by SAO in HEVC and is therefore delayed to be written
to the memory compared to H.264/AVC. Some additional memory is required for
keeping the SAO parameters (see Table
7.3
).
7.7
Conclusions
HEVC defines two in-loop filters, deblocking and sample adaptive offset (SAO),
which significantly improve the subjective quality of decoded video sequences
as well as compression efficiency by increasing the quality of the reconstruct-
ed/reference pictures. The deblocking filter attenuates discontinuities on the block
boundaries, while SAO mainly corrects ringing artifacts caused by large transforms
and quantization and sample value offsets in certain regions of a picture caused
by coding of motion vectors. The complexity of the HEVC deblocking has
been significantly reduced compared to the H.264/AVC. Moreover, the HEVC
deblocking filter is highly parallelizable with parallelization down to 8
8sample
blocks. Having lower computational complexity and being highly parallelizable,
the HEVC deblocking is less of a bottleneck in the decoder implementation than
the H.264/AVC deblocking and provides better trade-off between the computational
complexity and coding efficiency (subjective and objective quality). SAO is a new
in-loop filter, not present in H.264/AVC, which provides significant reduction of
ringing artifacts at relatively low decoding complexity. The deblocking and SAO
can also be implemented in the same processing unit, which simplifies CTU-based
encoding and decoding and reduces cost in hardware implementations.
References
1. Alshina E, Alshin A, Park JH (2012) AHG5: on bypass coding for SAO syntax elements, Joint
Collaborative Team on Video Coding (JCT-VC), Document JCTVC-J0043, Stockholm, July
2012
2. Alshina E, Alshin A, Park JH (2012) Encoder modification for SAO, Joint Collaborative Team
on Video Coding (JCT-VC), Document JCTVC-J0044, Stockholm, July 2012
3. Alshina E, Alshin A, Park JH, Fu C-M, Huang Y-W, Lei S (2012) AHG5/AHG6: on reducing
context models for SAO merge syntax, Joint Collaborative Team on Video Coding Coding
(JCT-VC), Document JCTVC-J0041, Stockholm, July 2012
4. Alshina E, Alshin A, Park JH, Laroche G, Gisquet C, Onno P (2012) AHG6: on SAO type
sharing between U and V components, Joint Collaborative Team on Video Coding (JCT-VC),
Document JCTVC-J0045, Stockholm, July 2012
5. Bjøntegaard G (2001) Calculation of average PSNR differences between RD-curves, ITU-T
SG16 Q6 Video Coding Experts Group (VCEG), Document VCEG-M33, Austin, Apr. 2001
