Joint Video Experts Team (jvet) of itu-t sg 6 wp and iso/iec jtc 1/sc 29/wg 11

[4]Project development (132)

Contributions in this category were discussed XXday XX July XXXX–XXXX (chaired by XXX).

4.1Text and general standard development (2)

JVET-K0263 VVC document structure [T. Amara (Amazon), D. Singer (Apple), A. Duenas (ARM), G. Martin-Cocher (BlackBerry), A. Hinds (CableLabs), T. Davies (Cisco), P. Pahalawatta (DirectTV), J. Samuelsson (Divideon), X. Ducloux (Harmonic), J. Chen (Huawei), J. Boyce (Intel), A. Norkin (Netflix), G. Teniou (Orange), M. Karczewicz (Qualcomm), J. Song (Samsung), T. Suzuki (Sony), D. Gibellino (Telecom Italia), D. Nicholson (Vitec)]

[Add summary – reference joint discussion.]

JVET-K0311 Interoperability point signalling for VVC [J. . Boyce (Intel), X. . Ducloux (Harmonic), A. . Hinds (CableLabs), S. . Wenger (Tencent)]

[Add summary – reference joint discussion.]

4.2Software development and CTC (10)

Discussion of the contributions on this subject was Aallocated to a BoG (coordinated by F. Bossen).
JVET-K0054 Unification of PSNR calculation for JVET CTC [R. . Chernyak, T. . Solovyev, S. . Ikonin, J. . Chen (Huawei)]
JVET-K0149 Reference software extension for coding block statistics [J. . Sauer, J. . Schneider, M. . Bläser, M. . Wien (RWTH Aachen)]
JVET-K0154 On encoding distortion evaluation of VTM/BMS software [T. . Chujoh, T. . Ikai (Sharp)]
JVET-K0496 Crosscheck report of JVET-K0154 on encoding distortion evaluation [S. . Iwamura, S. . Nemoto, A. . Ichigaya (NHK)] [late]
JVET-K0433 Crosscheck of JVET-K0154 on On encoding distortion evaluation of VTM/BMS software [X. . Li (Tencent)] [late]
JVET-K0261 AHG3: VVC Software Cleanup [A. . Wieckowski, K. . Sühring, H. . Schwarz]
JVET-K0312 AHG13: Reporting of adjusted decoder runtimes in tool on/off tests [J. . Boyce, D. . Gurulev, V. . Aristarkhov (Intel), A. . Tourapis (Apple)]
JVET-K0389 AHG3: Proposed software management for VTM [E. . Thomas, A. . Gabriel (TNO)] [late]
JVET-K0410 Comments on timing measurement variations for JVET experiments [F. . Glapin, T. . Poirier, F. . Le  Léannec, E. . François (Technicolor)] [late]

JVET-K0461 AHG3: Guidelines for VVC reference software development [F. . Bossen, X. . Li, K. . Sühring (AHG chairs)] [late]

4.3Common test conditions (1)

JVET-K0541 BoG report on common test conditions [J. Boyce]

This BoG met on 14 July 2018 (chaired by J. Boyce) to discuss forming recommendations about modifying the Class F sequences in the Common Test Conditions (CTC). The BoG had planned to meet again after some off-line work, but instead the further discussion took place directly in the JVET plenary.

Three eSports sequences were provided in JVET-K0294, all from the same video game. All of the sequences were somewhat similar. The first sequence, Tencent_AOV5 had the highest bit rate, and had some global motion. There was agreement in the BoG that it would be desirable to select one of the sequences to include in the new Class F, with a preference for the Tencent_AOV5 sequence.

An additional concert sequence provided with JVET-K0294 was viewed, but was not relevant for Class F.

The existing Class F sequences and the SCC Class TGM 4:2:0 sequences were viewed. The group expressed interest in replacing most of the sequences. Of the existing Class F sequences, SlideShow was considered the best. It is 720p resolution.

Several sequences from Twitch (no contribution uploaded yet) were informally viewed. All were 1080p60 4:2:0 and were eSports/video gaming content. It was verbally indicated that these sequences could be made available to the group.

The Twitch sequences have a lot of scene changes, but some sequences have reasonably long segments without scene changes, greater than 10 seconds. The sequences of most interest were called EuroTrucking (10 seconds near the end), and Witchers (early in the sequence).

It was suggested in the BoG to pick 10 second segments from each sequence and HEVC encode them, and to present the information in a late contribution.

It was suggested that some other SCC sequences may be of interest, including some that are 4:4:4, but could be converted to 4:2:0. MissionControlClip3 was particularly suggested, but was not available for viewing during the BoG meeting.

It was suggested in the BoG that it would be interesting to have text with both English alphabet and Asian characters.

Class F currently has 4 sequences, and there was consensus in the BoG to remain at 4 sequences, but to recommend changes to the sequences.

It was agreed in the BoG to aim for the following types of sequence content: 2 eSports, 1 desktop/powerpoint, 1 animation. The following were candidates for the 4 sequences:

• 
  eSports1: Tencent_AOV5 – first sequence from Tencent
  
• 
  eSports2: For 2nd sequence, consider the two sequences from Twitch, EuroTrucking and Witchers
  
• 
  Desktop/powerpoint: consider SlideShow from Class F, or MissionControlClip3 (4:4:4, would need to be converted to 4:2:0)
  
• 
  Animated: Look at sc_robot_1280x720_30_8bit, and other sources like sintel
  

As an action item for Tarek Amara, it was suggested to extract 10 second segments and encode the two possible clips from Twitch.

As an action item for Alberto Duenas for the Mission ControlClip3, to find the video and bit rates, and also to suggest some clips of animated content from sintel.

There was some discussion about whether Class F should have different QP values than the other sequences in the CTC, and whether the CTC generally should have higher QP values, to be more aligned with the CfP bit rate points.

Regarding other aspects of the CTC, there was some discussion in the BoG about what resolution should be used for the Class F content. The available content is at different resolutions, although downsampling is always possible. It was suggested for the same sequence at more than one resolution to be included in some test set(s), to help evaluate coding efficiency benefits at different resolutions.

It was suggested to possibly consider a new optional class with downsampled versions of sequences in other classes, so that the resolution impact could be easily measured. This issue was suggested to be raised in the JVET plenary, as it was considered not part of the BoG mandate.

There was a suggestion to use closed GOP for RA instead of open GOP, since closed GOPs are commonly used in adaptive streaming. There was also a suggestion to have a test condition defined with some type of scattered intra refresh, and perhaps to replace the LDP conditions with LDB with intra refresh.

This BoG report was discussed in the JVET plenary on Wednesday 1130 (chaired by GJS & JRO).

It was remarked that there could be a distinction between gaming/animation content and what we ordinarily consider screen content.

It was remarked that 4:4:4 content is desirable, and we should try to get 4:4:4 versions of sequences that we currently have as 4:2:0 if feasible.

It was suggested to replace ChinaSpeed with Tencent_AOV5 in Class F.

For future study, it was suggested to consider using some downsampled and higher-resolution sequences of the same source content and to consider having some content that is available in corresponding SDR and HDR versions.

It was commented that we should be cautious about changing Class F test sequences in order to avoid disrupting our ability to measure gain across meeting cycles. Others commented that since Class F had been optional and had not previously been emphasized in JVET, any change to Class F would not be disruptive.

Decision (CTC): The replacement of ChinaSpeed in Class F was agreed.

It was also discussed whether to make Class F mandatory in the CTC or not. No explicit decision was recorded in the meeting notes about that suggestion, so for purposes of preparation the meeting output documents it was assumed that no action was taken about that, such that Class F testing therefore remained optional.