Back to Search Document details
12th Meeting: Macao, October 2018 2018-10-03 00:39
CE1-related: On maximum/minimum allowed QT/BT/TT sizes for chroma
Authors: Chia-Ming Tsai MediaTek C.-W. Hsu MediaTek C.-Y. Chen MediaTek T.-D. Chuang MediaTek Y.-W. Huang MediaTek Shawmin Lei MediaTek
Abstract
When separate coding tree between luma and chroma is enabled in intra slices, maximum/minimum allowed QT/BT/TT sizes for chroma can be different from those derived from corresponding maximum/minimum QT/BT/TT sizes for luma according to chroma subsampling format. In this contribution, related parameters are named as MaxQtSizeC, MaxBtSizeC, MaxTtSizeC, MinQtSizeC, MinBtSizeC, and MinTtSizeC, where C stands for chroma. Results of four tests compared with VTM2.0.1 are reported as follows.
Meeting Notes
JVET-L0548 CE1-related: On maximum/minimum allowed QT/BT/TT sizes for chroma [C.-M. Tsai, C.-W. Hsu, C.-Y. Chen, T.-D. Chuang, Y.-W. Huang, S.-M. Lei (MediaTek)] [late]

When separate coding tree between luma and chroma is enabled in intra slices, maximum/minimum allowed QT/BT/TT sizes for chroma can be different from those derived from corresponding maximum/minimum QT/BT/TT sizes for luma according to chroma subsampling format. In this contribution, related parameters are named as MaxQtSizeC, MaxBtSizeC, MaxTtSizeC, MinQtSizeC, MinBtSizeC, and MinTtSizeC, where C stands for chroma. Results of four tests compared with VTM2.0.1 are reported as follows.

VTM2.0.1: {MaxQtSizeC, MaxBtSizeC, MaxTtSizeC, MinQtSizeC, MinBtSizeC, MinTtSizeC} in units of chroma samples = {32, 32, 16, 2, 2, 2}

Test1: Related parameters = {32, 32, 32, 2, 2, 2}
AI BD-rates = -0.08% (Y), -0.91% (U), -0.94% (V); EncT=103%; DecT=101%
RA BD-rates = 0.00% (Y), -0.32% (U), -0.31% (V); EncT=100%; DecT=100%
LB BD-rates = 0.01% (Y), -0.09% (U), -0.44% (V); EncT=100%; DecT=99%

Test2: Related parameters = {32, 32, 16, 4, 2, 2}
AI BD-rates = 0.00% (Y), 0.04% (U), 0.04% (V); EncT=100%; DecT=100%
RA BD-rates = 0.00% (Y), -0.01% (U), 0.00% (V); EncT=100%; DecT=100%
LB BD-rates = 0.00% (Y), 0.04% (U), -0.09% (V); EncT=100%; DecT=99%

Test3: Related parameters = {32, 32, 32, 4, 2, 2}
AI BD-rates = -0.08% (Y), -0.89% (U), -0.91% (V); EncT=103%; DecT=100%
RA BD-rates = -0.01% (Y), -0.36% (U), -0.30% (V); EncT=100%; DecT=100%
LB BD-rates = -0.02% (Y), -0.23% (U), -0.50% (V); EncT=100%; DecT=101%

Test4: Related parameters = {32, 32, 32, 4, 4, 4}
AI BD-rates = 0.06% (Y), 0.30% (U), 0.51% (V); EncT=100%; DecT=99%
RA BD-rates = 0.03% (Y), 0.55% (U), 0.78% (V); EncT=100%; DecT=100%
LB BD-rates = -0.05% (Y), 0.08% (U), 0.06% (V); EncT=100%; DecT=97%

It is suggested to Test1 or Test4 for consistent settings between QT, BT, and TT. It is also suggested that Test1 is slightly preferred than Test4 because of coding efficiency.

Related to restricting minimum block sizes. Tests 2..4 disallow 2-pixel sizes for chroma.

References:
JVET-K1010
Cited:
JVET-L0668 X
ARCHIVE v2 :
JVET-L0548-v1.docx JVET-L0548_Presentation.pdf JVET-L0548_Results_Table2.xlsm JVET-L0548_Results_Table3.xlsm JVET-L0548_Results_Table4.xlsm JVET-L0548_Results_Table5.xlsm
Decisions
Related to restricting minimum block sizes. Tests 2..4 disallow 2-pixel sizes for chroma.
Citation