Back to Search Document details
12th Meeting: Macao, October 2018 2018-10-07 05:21
CE3-related: One-line MMLM for reduction of reference sample lines
Authors: J. Lee KWU Joohyung Byeon KWU S. Park KWU Donggyu Sim KWU
Abstract
This contribution proposes a modified Multi Model Linear Model (MMLM) which uses one-line buffer for luminance in down-sampling process. The MMLM mode in BMS-2.0.1 with a macro as ‘--LMChroma=2’ is required four neighboring reconstructed luminance sample lines in the down-sampling process. The proposed method uses one reconstructed luminance samples line for above side in order to reduce line buffer. Experimental results show an average BD-rate gain 0.27%, 2.38% and 2.23% for three components on average over BMS-2.0.1 with vtm_config and ‘--LMChroma=2’, respectively.
Meeting Notes
JVET-L0066 CE3-related: One-line MMLM for reduction of reference sample lines [J. Lee, J. Byeon, S. Park, D. Sim (KWU)] [late]

This contribution proposes a modified Multi Model Linear Model (MMLM) which uses one-line buffer for luminance in down-sampling process. The MMLM mode in BMS-2.0.1 with a macro as ‘--LMChroma=2’ is required four neighbouring reconstructed luminance sample lines in the down-sampling process. The proposed method uses one reconstructed luminance samples line for above side in order to reduce line buffer. Experimental results show an average BD-rate gain 0.27%, 2.38% and 2.23% for three components on average over BMS-2.0.1 with vtm_config and ‘--LMChroma=2’, respectively.

Same concept as JVET-L0065. The comparison is made for MMLM with 4 lines for computing, therefore the loss is higher. As MMLM is not adopted, no action necessary.

Cited:
JVET-L0023 JVET-M0423 X
ARCHIVE v5 :
._.DS_Store ._exp ._JVET-L0065 & L0066_v2.pdf ._JVET-L0066-v1.docx ._JVET-L0066-v2.docx ._JVET-L0066-v3.docx ._JVET-L0066-v4.docx ._JVET-L0066-v5 ._JVET-L0066_TEST1_BMS201_vtm_+CbtoCr+MMLM.xlsm ._JVET-L0066_TEST2_BMS201_vtm_+CbtoCr+Oneline_MMLM.xlsm ._JVET-L0066_TEST3_BMS201_vtm+MMLM.xlsm ._JVET-L0066_TEST4_BMS201_vtm+One_line_MMLM.xlsm .DS_Store JVET-L0065 & L0066_v2.pdf JVET-L0066-v1.docx JVET-L0066-v2.docx JVET-L0066-v3.docx JVET-L0066-v4.docx JVET-L0066_TEST1_BMS201_vtm_+CbtoCr+MMLM.xlsm JVET-L0066_TEST2_BMS201_vtm_+CbtoCr+Oneline_MMLM.xlsm JVET-L0066_TEST3_BMS201_vtm+MMLM.xlsm JVET-L0066_TEST4_BMS201_vtm+One_line_MMLM.xlsm
Decisions
adopted
Same concept as JVET-L0065. The comparison is made for MMLM with 4 lines for computing, therefore the loss is higher. As MMLM is not adopted, no action necessary.
Citation