n0444:VVC/VTM:第14次会议接收提案汇总 2024-04-24 14:09:35 0 0 JVET-N0671:Support of 4:4:4 and 4:2:2 chroma formats in VVC 为了支持4:4:4 和 4:2:2 格式对CCLM和Affine进行修改。 JVET-N0266 test 2:CE2-related: Disabling bi-prediction or inter-prediction for small blocks-test2 • Disable inter prediction for 44 CUs and disable bi-prediction for 48 and 8*4 CUs: there is 0.08% and 0.11% coding loss in RA and LDB Main10 configurations respectively. JVET-N0068:with a max difference of 2 integer positions JVET-N0271:CE3-1.5: CCLM derived from four neighbouring samples JVETN0137, test 3-2.1.1:CE3-2.1: Intra chroma partitioning and prediction 方法: Disable 2x2/4x2/2x4 in dual tree only 目的: alleviate intra chroma throughput issues(separate tree中帧内的色度块的处理pipeline太长); resolving the dependency issues of small chroma blocks. ??? JVET-N0217:CE3: Affine linear weighted intra prediction (CE3-4.1, CE3-4.2)???? JVET-N0324:CE4: Regular merge flag coding (CE4-1.2.a and CE4-1.2.b) Test 4-1.2:The modified signalling order is: Regular merge, MMVD, subblock, CIIP, and TPM, rather than MMVD, subblock, CIIP, TPM, and then regular merge. show -0.01% and -0.07% BD-rate improvements under RA and LDB, respectively JVET-N0242: version 5-3.2 JVET-N0415:,version 5-4.2 JVET-N0415:version 5-4.2 JVET-N0193:version from CE6-3.1d. CTC for AI and RA, not LB. JVET-N0054: JVET-N0188: JVET-N0280: method 8-4.4a, counter-based limitation with 2 context coded bins. It was later reported by the cross-checker that there were initially some misalignments between text and software, which were however resolved in the upload JVET-N0280v3. JVET-N0407:CE-9-1.1a JVET-N0178: CE9-2.4 JVET-N0302:CE10-1.1 JVET-N0340:CE4-related JVET-N0866: JVET-N0103-V3 JVET-N0492:CBF flags signalling in VVC,去除hierarchical chroma CBF的冗余。 JVET-N0847 JVET-N0473 JVET-N0246, method 6 JVET-N0600 method 2 JVET-N0220 aspects 2,3, and 4 …好像太多了,有很多小的idea,先写到这里吧,后面再整理这次会议接收的重要提案。 根据最新的标准文档JVET-N1001得到下面的主要接收提案 • JVET-N0067: NAL unit header. • JVET-N0278: Layer concept without inter-layer referencing. • JVET-N0349: Decoding parameter set (DPS). • JVET-N0869: Semantics of DPS ID equal to 0. • JVET-N0805: APS types (with loop). • JVET-N0100: Signalling information about long-term reference picture POC LSB. • JVET-N0100: Conditional signalling of num_ref_idx_active_overwrite_flag. • JVET-N0288: Inference rule for num_bricks_in_slice_minus1. • JVET-N0124: Inference rule for single_brick_per_slice_flag and signalling the bottom right brick index as a delta from top left brick index. • JVET-N0047: Ref pic list for IDR. • JVET-N0276: 5 new constraint flags. • JVET-N0276: Sub-profile indication. • JVET-N0865: Gradual random access (GRA). • JVET-N0101: CRA decoding process. • JVET-N0070: Ref pic wraparound cleanups. • JVET-N0352: Conformance window. • JVET-N0438: Loop filtering disabling aross virtual boundaries. • JVET-N0498: Uniform tile partitioning. • JVET-N0150: WPP with 1-CTU lag. • JVET-N0120: Editorial improvements to the general decoding process to clarify the difference between a CVS and a bitstream. • JVET-N0857: Tile and brick partitioning. • JVET-N0706: Decoded picture hash SEI message. • JVET-N0494: Dependent RAP indication SEI message. • JVET-N0353: Buffering period & picture timing SEI messages. • JVET-N0867: Temporal scalability HRD parameters. • JVET-N0063: VUI design. • JVET-N0423 & N0350: HRD starting point. • JVET-N0266: Remove 4x4 unipred, and 4x8/8x4 bipred regular inter modes, and remove the use of for shared merge candidates in the regular merge list. • JVET-N0821: 6-tap interpolation filter for affine MC. • JVET-N0068: Restriction of memory bandwidth consumption of affine MC. • JVET-N0481: BCW index inheritance for constructed affine merge candidate. • JVET-N0334: MV clipping for MMVD, DMVR and CPMV of constructed affine merge candidadates. • JVET-N0407: Disable 8x8/4xN CUs for DMVR. • JVET-N0178: Implicitly split BDOF application region along 16x16 boundaries. • JVET-N0146: Disabling BDOF for BCW and WP, and align the condidtion of DMVR and BIO. • JVET-N0444: Condition check of block height is not equal to 4 for BDOF. • JVET-N0325: 8-bit fixed precision for BDOF calculations. • JVET-N0127: MMVD enabling signalling in SPS • JVET-N0332: disable MVD scaling for MMVD for LTRP. • JVET-N0448/JVET-N0380: infer the number of MMVD base merge candidate to 1. • JVET-N0302: CIIP with position-independent weights. • JVET-N0324: Signal a regular merge flag right after the merge flag and skip flag. • JVET-N0851/N0447/N0400/N0500: align triangle merge candidate number and regular merge candidate number. • JVET-N0340: regular merge candidate list is re-used for triangle. • JVET-N0483: disable sub-block transform when triangle mode is used. • JVET-M0335: Round MVs toward zero. • JVET-N0235: Add an SPS-level flag to turn on and off symmetric MVD • JVET-N0213: Remove TMVP from merge and AMVP list for 4x8 and 8x4 CUs • JVET-N0868: Revised spec text for DMVR reconciling with software ticket #214 收藏(0)