Back to Search
Document details
AHG9: Constraints based on the minimum coding block size
Abstract
In the current VVC text, there are some cases that assertedly cause contradition or confusion considering MinCbSizeY:
JVET-R0237 AHG9: Constraints based on the minimum coding block size [K. Zhang, L. Zhang, Y.-K. Wang, Z. Deng, Y. Wang, J. Xu, H. Liu (Bytedance)]
In the current VVC text, there are some cases that assertedly cause contradition or confusion considering MinCbSizeY:
- The maximum allowed block size for transform skip may be smaller than MinCbSizeY.
- log2_parallel_merge_level_minus2 may be smaller than log2_min_luma_coding_block_size_minus2.
- The coding tools MTS, LFNST, SBT, ISP can only be applied to blocks smaller than a size, which may be smaller than MinCbSizeY.
Four constraints are proposed in this contribution to address the problems above:
- It is constrained that the maximum allowed block size for transform skip cannot be less than MinCbSizeY.
- It is constraend that log2_parallel_merge_level_minus2 must be greater than or equal to log2_min_luma_coding_block_size_minus2.
- It is constrained that sps_transform_skip_enabled_flag and sps_mts_enabled_flag must be equal to 0 when MinCbSizeY is greater than 32.
- It is constrained that sps_lfnst_enabled_flag, sps_sbt_enabled_flag, and sps_isp_enabled_flag must all be equal to 0 when MinCbSizeY is greater than MaxTbSizeY.
The constraint 2) is trying solving the same inconsistency as JVET-R0216. It prevents signalling values that would be overridden somehow later, and therefore the syntax is redundant.
It is however pointed out that there is no benefit for the decoder. Several experts expressed concerns if these specific changes would even give benefit for an encoder.
No action was taken on this.
References:
PATENTS:
TWI771590B
0.34
2022-07-21
20230046629
0.34
20230021488
0.33
US8023757B2
0.32
2011-09-20
20230027478
0.32
20240137570
0.31
US20250287046A1
0.30
2025-09-11
4307674
0.30
Decisions
No action was taken on this.
Citation