Tags give the ability to mark specific points in history as being important
  • VTM-3.0rc1   VTM version 3.0 release candidate 1

    Includes following items:

    • L0082 10 b coeffs (instead of 11) – this seems clearly the logical thing to do, no loss
    • L0147 Adopt 2.6.2, CE2: Subsampled Laplacian calculation
    • L0083 Adopt (text is in the contribution), JVET-L0083, Reduction of bits for ALF coefficient fractional part
    • L0392 Decision (minor BF): Adopted
    • L0664 Remove the alf_luma_type_flag and the conditioning on it that results in signalling of 5x5 as a special case for luma
    • L0283 CE3: Multiple reference line intra prediction, version 1.1.4 (with line restriction from CTU above).
    • L0628 CE 3.1.4.2 A combination of tests 3.1.2 and 3.1.4 for intra reference sample interpolation filter
    • L0191 CE3: Cross-component linear model simplification (Test 5.1)
    • L0136/L0085 L0136 (5.2.7) / JVET-L0085 (5.8.1) CCLM with line buffer restriction
    • L0338/L0340 L0338 method 5.4.1/JVET-L0340 method 5.6.1 Multi-directional LM
    • L0053/L0272 Adopt JVET-L0053 first aspect / JVET-L0272 (CE3-related: Chroma DM modification)
    • L0279 CE3-related: Unification of angular intra prediction for square and non-square blocks
    • L0165 CE3-6.2.1: Extended MPM list
    • L0059 CE6-related: Simplification on MTS kernel derivation
    • L0111 CE6-related: Transform Skip Condition on Transform Block size
    • L0285 Adopt JVET-L0285 (8-bit transform matrices)
    • L0118 CE6-1.7b: Currently the switching DST-7/DCT-8 is different for inter and intra. Results indicate that the unification has minor impact on performance
    • L0274 version 7.1.3b Transform coefficient coding with reduced number of regular-coded bins
    • L0414 Other from the proposal, which makes the QP offset dependent on transfer function, the values shall be signalled in the SPS. Default is not applying (enabling flag=0). If the flag is 1, another syntax element follow indicating the number of intervals (2 bits for 2,3,4,5), and then the luma threshold values and QP offsets between the intervals.
    • L0410 CE11-related: On deblocking tC table Meeting notes (L0074) Apply the same logic to VVC (both ATMVP and affine) sub-blocks (on 8x8 grid) as to PU in HEVC deblocking. This means check the deblocking motion conditions for ATMVP and affine motion sub-block boundaries as if they were PUs in HEVC.
    • L0217 Adopt JVET-L0217 (as per v4), Non-CE1: Relation Between QT/BT/TT Split Constraint Syntax Elements
    • L0678 QT/BT/TT Split Constraint Syntax Elements Signaling Method
    • L0081 CE1.2.1: Constraint for binary and ternary partitions
    • L0361 JVET-L0361 (version with 22 context models) contexts for split
    • L0553 Adopt second fix to semantics of init_qp_minus26 where +25 is changed to +37
    • L0293 Adopt CPR with restriction to using current CTU as reference area under condition, non-CTC
    • L0047 Adopt method 1 as the more consistent and “clean” design (roughly neutral on coding efficiency 0.01%
    • L0198, L0468, L0104 "use fixed subblock size 8x8 for ATMVP (L0198, L0468, L0104, possibly some others). check the first spatial neighbouring motion vector and use this as the reference motion vector for the collocated position for motion vector derivation "
    • L0106, L0158 Reset the FIFO table in each CTU row for HMVP
    • L0194 use one context for the first bin of the full-block merge index and bypass coding of other bins . for the full-block merge index only at this time, approximately no coding efficiency impact
    • L0646 Adopt L0646 (0.66% coding efficiency; weighted prediction should also be put in the draft, but this and weighted prediction would be mutually exclusive at the picture level, when used with OBMC the weights of the neigbours would apply for the neighbour predictors, which is how the BMS software already does it, no consideration in deblocking filter
    • L0104 Prohibit 4x4 bi-prediction for inter CU
    • L0257 Align the software with the draft text regarding ATMVP motion vector clipping
    • L0377 Rounding motion vectors toward zero rather than toward minus infinity for AMVR
    • L0093/L0282 align VTM with draft text regarding the pruning of regular merge list
    • L0265 set the chroma subblock size to 4x4 instead of 2x2 for affine motion compensation by averaging the MVs of the 4x4 luma subblocks
    • L0271 Adopt 4.1.6.a (CE4.1.6: Simplification of affine AMVP candidate list construction)
    • L0045 Adopt 4.1.11.a on line buffer reduction for affine mode (pending consideration of non-CE contributions, text in JVET-L0045 to be checked).
    • L0632/L0142 Adopt the variation of 4.2.6.d as modified in L0632, pending text adapted from modifying model inheritance of L0368 and cross-check confirmation
    • L0369/L0055 Adopt 4.2.8 moving ATMVP into the affine merge list (assuming ATMVP operates on a subblock basis, restrict ATMVP mode to CUs of which both the width and height are larger than or equal to 8 (L0055)
    • L0090 4.4.12.a (0.38% in RA), merge list size 6 (text in L0090)
    • L0266 Adopt history method with merge list size 6, history applied also to AMVP (0.58% gain in RA, text in L0266).
    • L0054 Adopt UMVE variant b (1.29% in RA); this needs a better name – e.g., merge with MVD (MMVD)
    • L0694 CE4-related: Combination of affine mode clean up and line buffer reduction
    • L0260 Encoder optimization for affine motion estimation
    • L0100 Adopt CE10.1.1.c (described in JVET-L0100) combined intra/inter with restriction to w×h >= 64 luma samples (0.5% in RA).
    • L0124/L0208 CE10.3.1.b, Adopt (0.57% in RA, 1.23% in LB), with the L0208 bug fix, flag after combined intra/inter
    • L0256 CE9-related: Complexity reduction and bit-width control for bi-directional optical flow (BIO)
    • L0552 CABAC initial state retraining
  • VTM-2.2   VTM version 2.2
  • VTM-2.1   VTM version 2.1
  • VTM-2.0.1   VTM version 2.0.1

    Bug fixes:

    • uninitialized memory access in intra prediction (bug #77)
    • uninitialized memory access in affine ME (bug #76)
  • VTM-2.0   VTM version 2.0
    39b913ea · missing dot-files ·

    Added features for version 2.0:

    • K0054: high-precision PSNR reporting
    • K0063: PDPC
    • K0072: trellis coded quantization
    • K0122: DC prediction
    • K0154: high-precision distortion
    • K0184: Affine MC
    • K0190: CCLM
    • K0220: Encoder speedup
    • K0230: Dual coding tree
    • K0251: extended QP range
    • K0261: SW cleanup
    • K0346: ATMVP
    • K0351: TT restriction
    • K0352: Encoder optimization of merge
    • K0357: AMVR
    • K0371: ALF
    • K0312: additional decoder stats
    • K0238: SAO greedy merge encoder option
    • K0500: wide angular intra prediction
    • K0554: boundary handling
    • K0556: maximum TT size is 64
    • K1000: simplified EMT
    • 67 intra modes with 3 MPMs
    • Various deblocking fixes
    • -ipp options now adds a / in path if needed
    • Increase chroma QP with dual coding trees are used
    • Remove type aliases such as Int, Void, etc.
    • Updated license text (include year 2018)

    Fixes since rc1:

    • Separate tree fix
    • Remove unused code (VTM)
    • Update high-precision PSNR reporting to also apply to 360 case
    • Fix compilation issue when tracing is enabled
    • Fix chroma lambda when Cb to Cr prediction is off
  • VTM-2.0rc2   VTM version 2.0 release candidate 2
  • VTM-2.0rc1   VTM version 2.0 release candidate 1