Search Results for "JVET-L0410"

Found 4 document(s)

Search documents

Use number, keyword, author, or MPEG number. Filter by meeting when needed.

12th Meeting: Macao, October 2018 2018-10-07 06:30
Abstract
The document argues that the values chosen during the last meeting to extend the deblocking tC table (as a consequence of extending the maximum QP value to 63) do not reflect the intent of the tC table design. The document proposes new values to extend the tC table that keep linear relation of tC with the quantization step size which is asserted to be the design intent for the deblocking tC table.
JVET-L0410 CE11-related: On deblocking tC table [A. Norkin (Netflix)]

Decision: Adopt JVET-L0410, updated tC table into VVC text & VTM (Track A Wed 10 Oct., 1200, chaired by JRO)

Decisions
adopted
Adopt JVET-L0410, updated tC table into VVC text & VTM (Track A Wed 10 Oct., 1200, chaired by JRO)
Citation
12th Meeting: Macao, October 2018 2018-11-27 21:56
Abstract
The main goal of this Core Experiment (CE) is to study removing subjective artifacts at face boundaries, comparing 360° specific coding tools with existing coding tools, such as tiles, combined with non-normative pre- and post-processing. The tested 360° specific coding tools include modifications to in-loop filtering and inter prediction, as well as projection format adaptation. To alleviate seam artifacts, the tested solutions shall modify in-loop filters by either disabling filtering across face discontinuities and/or by using spherical neighbors in the filtering process. The tested solutions will be implemented for the hybrid equi-angular cubemap (HEC) projection and compared to two anchors. The first anchor is the 360° CTC anchor (JVET-L1012), i.e., HEC with padding of 4 samples around face row with blending (PHEC). The second anchor is the CE anchor, which is PHEC with in-loop filters disabled across face discontinuities.
JVET-L1033 Description of Core Experiment 13 (CE13): Coding tools for 360° omnidirectional video [P. Hanhart, J.-L. Lin, C. Pujara]

This CE was planned to study changes to in-loop filters, inter prediction, post-filtering, projection rotations and packings.

Future meeting plans, expressions of thanks, and closing of the meeting

Future meeting plans were established according to the following guidelines:

  • Meeting under ITU-T SG 16 auspices when it meets (starting meetings on the Tuesday or Wednesday of the first week and closing it on the Tuesday or Wednesday of the second week of the SG 16 meeting – a total of 6–7.5 meeting days), and
  • Otherwise meeting under ISO/IEC JTC 1/SC 29/WG 11 auspices when it meets (starting meetings on the Wednesday or Thursday prior to such meetings and closing it on the last day of the WG 11 meeting – a total of 8.5 meeting days).

In cases where high workload is expected for a meeting, an earlier starting date may be defined.

Some specific future meeting plans (to be confirmed) were established as follows:

  • Wed. 9 – Fri. 18 January 2019, 13th meeting under WG 11 auspices in Marrakech, MA.
  • Tue. 19 – Wed. 27 March 2019, 14th meeting under ITU-T auspices in Geneva, CH.
  • Wed. 3 – Fri. 12 July 2019, 15th meeting under WG 11 auspices in Gothenburg, SE.
  • Tue. 1 – Wed. 9 October 2019, 16th meeting under ITU-T auspices in Geneva, CH.

The agreed document deadline for the 13thJVET meeting was planned to be Wednesday 2 January 2019 (a little later than usual due to the proximity of the New Year holiday). Plans for scheduling of agenda items within that meeting...

Decisions
Jian Qing Zhu (Fujitsu R&D Center)
Citation
12th Meeting: Macao, October 2018 2018-10-11 17:47
Abstract
The BoG on Neural Networks for Video Coding was held on 11 October 2018. The BoG discussed several subjects including: training data set and process, software sharing and simulation, neural network structure and complexity analysis, reporting and comparison procedure, etc. The BoG recommends:
JVET-L0704 BoG report on Neural Networks for Video Coding [S. Liu, Y. Li]

The BoG on Neural Networks for Video Coding met on 11 October 2018. The BoG discussed several subjects including: training data set and process, software sharing and simulation, neural network structure and complexity analysis, reporting and comparison procedure, etc. The BoG recommended:

  • To use the Div2K set as the base training set for all proposals;
  • To establish a software branch for neural network coding tools (proposals);
  • A template to report neural network structure and complexity;
  • To consult with hardware experts about practical complexity criteria;
  • To continue discussing related issues in an ad-hoc group.

The BoG outcome was reviewed Fri 12 Oct 0930 (GJS).

It was noted that the Div2K contains only still images (and only 800 of them). Its use was only encouraged in order to ease comparisons if applicable. Contributors may use other data for training, but need to describe what they did.

No particular training platform was selected, but the method should be described, and the amount of time spent on the training should be described.

Among the proposed technologies at this meeting, the training was always something that had been done offline, not as part of the sequence coding/decoding process.

It was clarified that the study is encouraged to include not only proposed normative elements that would affect the decoding process, but also other aspects – e.g., encoder-only processing (for mode decisions, partitioning, etc.) and pre-/post-processing.

There had been some discussion of potential...

Decisions
A template had been developed and was attached to the BoG report for use in reporting of results and analysis. Its use in further experiments was encouraged.
adopted
Adopt.Regarding the question of whether we have a PPS or picture header or both.Decision: Use a PPS for now.The picture header topic as a substitute or additional header is for further study.Whether to have some picture header is primarily a coding efficiency issue.Decision: Regarding including some provision for future extensions in the syntax (e.g., multi-layer extension), it was agreed that may do so in cases where this has a very minor impact on the current design
Citation
12th Meeting: Macao, October 2018 2018-10-11 08:29

Abstract not available in document

JVET-L0681 BoG report on CE11: Deblocking filter [A. Norkin, A. Segall, A. M. Kotra (Huawei)] In the Track A meeting Mon. 8 Oct. 1400, it was agreed to merge the previous two planned BoGs on deblocking, and nominate A. Norkin, A. Segall, A. Kotra as chairs of that BoG, and extend the mandates to perform further analysis of CE11 contributions, and also review the contributions from 7.11, and recommend items to be investigated in the upcoming CE11. The Breakout Group on Deblocking met during the 12th meeting. The Breakout Group was originally created as two Breakout Groups. The first on Long Tap Deblocking Filters, and the second on Deblocking for 4 x N and N x 4 block boundaries. Each of these BoGs met separately on October 7th, and then were combined on October 8th. The goal of the combined group was as follows: Long Tap Deblocking Further analyze the results of subjective test in JVET-L...
Decisions
The recommendation of the BoG was agreed in JVET. For CE work on any other subblock-base modes (planar MV mode and others if any), the same scheme should apply (at least as an anchor)
See section 6.11.
Citation
New Search