International organisation for standardisation

6.8.2MV Coding

6.8.2.1.1.1.1.1.1JCTVC-F105 On MVP list pruning process [Y. Jeon, B. Jeon (LGE)]

Discussed in BoG coordinated by B. Bross.

6.8.2.1.1.1.1.1.2JCTVC-F086 Cross verification of LGE's proposal JCTVC-F105 on MVP pruning process [M. Zhou (TI)]
6.8.2.1.1.1.1.1.3JCTVC-F099 Performance report of temporal motion vector prediction [H. Aoki, K. Chono, Y. Senda (NEC)]

Discussed in BoG coordinated by B. Bross.

6.8.2.1.1.1.1.1.4JCTVC-F374 Cross-verification report of JCTVC-F099 by JVC Kenwood [S. Fukushima (JVC Kenwood)] [late upload 07-05]
6.8.2.1.1.1.1.1.5JCTVC-F142 Improvement on derivation process for luma motion vector prediction [K. Kazui, S. Shimada, J. Koyama, A. Nakagawa (Fujitsu)]

This contribution proposed two modifications of the derivation process for luma motion vector prediction in WD3.0. Firstly, the rounding process in the equations for scaling motion vector was modified. Secondly, the derivation process for temporal luma motion vector prediction was modified so that a motion vector predictor is chosen among two motion vectors in a collocated picture.

The coding gain of the proposed scheme for BQSquare sequence is reportedly up to 2.0%. The average coding gain for Class D sequences for RA HE and RA LC is reportedly 0.9%. The overall coding gains for random access configurations and low delay configurations are reportedly 0.5% and 0.1%, respectively, for both HE and LC. Additional computational complexity both in encoding and decoding is reportedly negligible.

This has two modifications, but most of the gain is from one of them.

That modification is to round toward zero when applying MV scaling.

Decision: Adopt this aspect.

The other part was not adopted.

6.8.2.1.1.1.1.1.6JCTVC-F183 Cross-verification report on Fujitsu's proposal (JCTVC-F142) on AMVP [K. Sugimoto, A. Minezawa, S. Sekiguchi (Mitsubishi)]

6.8.2.1.1.1.1.1.7JCTVC-F373 Merge based MVD transmission [S. Fukushima, M. Ueda, K. Arakage, S. Sakazume (JVC Kenwood)]

Discussed in BoG coordinated by B. Bross.

6.8.2.1.1.1.1.1.8JCTVC-F427 Consideration on Temporal Predictor [K. Sato (Sony)]

Discussed in BoG coordinated by B. Bross.

6.8.2.1.1.1.1.1.9JCTVC-F626 Cross-verification result of JCTVC-F427 proposed by Sony [K. Kazui (Fujitsu)] [late reg. 07-04, upload 07-04]
6.8.2.1.1.1.1.1.10JCTVC-F575 Simplification of MVP Design for HEVC [Y. Yu, K. Panusopone, L. Wang (Motorola Mobility)]

Discussed in BoG coordinated by B. Bross.

6.8.2.1.1.1.1.1.11JCTVC-F587 Reduction of reference picture list checking for temporal motion vector prediction [I.-K Kim, W.-J Han, JH Park (Samsung)]

Discussed in BoG coordinated by B. Bross.

6.8.2.1.1.1.1.1.12JCTVC-F657 Cross verification of Samsung's proposal JCTVC-F587 on Reduction of reference picture list checking for temporal motion vector prediction [M. Zhou (TI)] [late reg. 07-05, upload 07-07]
6.8.2.1.1.1.1.1.13JCTVC-F465 Item 1 (Center position of TMVP candidates) of Experiments on tools in Working Daft (WD) and HEVC Test Mode (HM-3.0) [I.-K. Kim, E. Alshina, J. Chen, T. Lee, W.-J. Han, J. H. Park (Samsung), V. Sze, M. Budagavi, M. Zhou (TI)]

Item 1: Center position of TMVP candidates

Discussed in BoG coordinated by B. Bross – will be tested in CE9.

6.8.2.1.1.1.1.1.14JCTVC-F081 CE1: Evaluation results on A.09, A.13-16 and an alternative solution [M. Zhou (TI)]

This contribution is relevant to CE1, and is discussed above in that context, but also contains a new proposal. The new proposal is the same as (or strongly similar to) item 1 of JCTVC-F465, and was asserted to be suitable as a cross-check of that proposal. Placed under consideration in BoG coordinated by B. Bross.

6.8.2.1.1.1.1.1.15JCTVC-F341 Decoder Performance Restrictions due to Merge/MVP Index Parsing [T. Hellman, Y. Yu (Broadcom)]

Discussed in parsing robustness BoG coordinated by B. Bross.

6.8.2.1.1.1.1.1.16JCTVC-F372 Bi-derivative merge candidate [H. Takehara, S. Fukushima, H. Nakamura (JVC Kenwood)]

In merge mode, motion information used for motion compensation is indicated only by signalling the index of merge candidates. The merge candidates are discussed in this contribution in cases where some candidates are not available.

In this proposal, additional merge candidates are created by using motion information of current merge candidates without additional access to other blocks.

It was reported that the proposed technique provides 0.2% BD-rate gain for RA settings and 0.7% gain for LB settings with 3% encoder runtime increase and 1% decoder runtime increase.

The encoder runtime increase is due to an increase in the average number of merge candidates to be evaluated.

The same concept is reportedly proposed in JCTVC-F470.

Should be studied in relation to the MVP candidate list construction discussion. Discussed in BoG coordinated by B. Bross.

6.8.2.1.1.1.1.1.17JCTVC-F419 Unification of derivation process for merge mode and MVP [H. Nakamura, S. Fukushima, M. Nishitani (JVC Kenwood)]

This contribution presents simplifications of derivation process for merge mode and motion vector predictor (MVP). In this proposal, the positions of the spatial neighbors that can be used as merging candidates are as same as the positions of the spatial MVP candidates. That is unification of the location of spatial neighbors for merge mode and MVP. In addition, the proposed techniques attempt unification of the derivation process for merge mode and MVP. This proposal tries to reduce the number of candidates in the spatial derivation process to reduce the number of times of comparison in the removal process.

The proposed technique reportedly provides 0.1-0.2% BD-rate loss for random access, and 0.0-0.2% loss for low delay.

Two sub-proposals are in the contribution. Discussed in BoG coordinated by B. Bross.