Back to Search Document details
12th Meeting: Macao, October 2018 2018-10-06 04:58
CE11-related: CTU line buffer reduction for long filter deblocking
Authors: Anand Meher Kotra Huawei S. Esenlik Huawei Biao Wang Huawei Jie Chen Huawei Weijia Zhu Sharp Kiran Misra Sharp P. Cowan Sharp Andrew Segall Sharp
Abstract
The contribution is a joint proposal between Huawei and Sharp. Long filter deblocking is asserted to provide significant visual benefit over the current approach in VTM-2.0.1. However, such approaches also extend the number of CTU line buffers. The contribution proposes modifying the filtering and control processes at the top edge of the CTU boundary to access samples within 4 luma lines and 2 chroma lines. Aspects of long filter deblocking and CTU line buffer aware filtering are based on tests CE11.1.7b and CE11.1.8. The proposal includes two further modifications, one for subsampling the chroma gradient computation to further reduce complexity and a threshold modification to improve objective gains. It is asserted that the visual impact of these changes is negligible. Objective evaluations show a luma BD rate impact of -0.02%/ 0.03%/0.15%,0.10%for AI/RA/LD/LDP compared to VTM-2.0.1 and -0.03%/0.02%,0.17%,0.03% for AI/RA/LD/LDP compared to VTM-2.0.1 when ALF is turned off.
Meeting Notes
JVET-L0572 CE11-related: CTU line buffer reduction for long filter deblocking [A. M. Kotra, S. Esenlik, B. Wang, J. Chen (Huawei), W. Zhu, K. Misra, P. Cowan, A. Segall (Sharp)] [late]
References:
JVET-K1002 JVET-K1010 JVET-K1031 JVET-L0224 JVET-L0405
Cited:
JVET-L0661 X JVET-L1031 JVET-M0031
PATENTS:
CN120499372A 0.71 2025-08-15 KR20200124738A 0.32 2020-11-03 CN117319674A 0.30 2023-12-29 CN114514747A 0.28 2022-05-17 CN114424527A 0.26 2022-04-29 CN116744002A 0.24 2023-09-12 CN114467116A 0.22 2022-05-10 JP2023160851A 0.20 2023-11-02
ARCHIVE v2 :
JVET-L0572-v2.docx JVET-L0572-v2.pptx JVET-L0572_v2.xlsm JVET-L0572_v2_noALF.XLSM
Decisions
JVET-L0572 CE11-related: CTU line buffer reduction for long filter deblocking [A. M. Kotra, S. Esenlik, B. Wang, J. Chen (Huawei), W. Zhu, K. Misra, P. Cowan, A. Segall (Sharp)] [late]
Citation