Back to Search Document details
15th Meeting: Gothenburg, SE, July 2019 2019-07-07 16:24
Non-CE5: Unified design for longer tap deblocking line buffer reduction
Abstract
In VTM-5.0, to reduce line buffer requirement for the longer tap filter at horizontal CTU boundaries, for the Luma component an asymmetric filter which only modifies 3 samples from the top CU, and 7 samples from the bottom CU is employed, whereas for the Chroma deblocking, the longer tap filter is completely switched off at the CTU boundaries. Since a potential “large block” blocking artifact could occur at the horizontal Chroma CTU boundary, the current proposal reportedly employs a longer tap asymmetric filter which modifies 1 sample from the top CTU but can still modify up to 3 samples from the bottom CTU to more effectively remove the blocking artifact.
JVET-O0803 Non-CE5: Unified design for longer tap deblocking line buffer reduction [A. M. Kotra, S. Esenlik, B. Wang, H. Gao, E. Alshina (Huawei)] [late]

In VTM-5.0, to reduce line buffer requirement for the longer tap filter at horizontal CTU boundaries, for the Luma component an asymmetric filter which only modifies 3 samples from the top CU, and 7 samples from the bottom CU is employed, whereas for the Chroma deblocking, the longer tap filter is completely switched off at the CTU boundaries. Since a potential “large block” blocking artefact could occur at the horizontal Chroma CTU boundary, the current proposal reportedly employs a longer tap asymmetric filter which modifies 1 sample from the top CTU but can still modify up to 3 samples from the bottom CTU to more effectively remove the blocking artefact.

The objective results, over VTM5.0 Anchor for CTC configuration are as follows

Config.

Y

U

V

EncT

DecT

AI

0.00%

0.00%

0.00%

101%

101%

RA

0.00%

-0.06%

-0.04%

101%

102%

LDB

0.02%

0.00%

-0.10%

101%

102%

LDP

0.00%

0.18%

-0.05%

101%

102%

Subjective evaluation showed that the quality is slightly better when compared to the Anchor and reportedly the subjective benefits are better for 4:2:2 and 4:4:4 content.

Further study in CE. Unification is not necessary an argument, as there is a slight complexity increase, and should be justified by evidence that artefacts are avoided.

Decisions
Further study in CE. Unification is not necessary an argument, as there is a slight complexity increase, and should be justified by evidence that artefacts are avoided.
Citation