Joint Collaborative Team on Video Coding (jct-vc) Contribution

6.7Withdrawn and unclear allocation

JCTVC-M0398 Multiview and multiview-video-plus-depth support in SHM [A. Hallapuro, Y. Yan, D. Rusanovskyy, M. M. Hannuksela] [late]
JCTVC-M0038 [Withdrawn]
JCTVC-M0050 [Withdrawn]
JCTVC-M0248 [Withdrawn]
JCTVC-M0332 [Withdrawn]
JCTVC-M0227 [Withdrawn]
JCTVC-M0228 Cross-check Non-SCE [P. Lei (Samsung)] [miss]

7Plenary Discussions and BoG Reports

7.1Project development

Report of tracks A and B Sunday morning plenary

Many tools were considered under aspects of complexity, for which a more consistent procedure had been developed during the last meeting. In terms of complexity assessment, the current SHM decoder in worst case could be equal or more complex than the sum of parallel operated simulcast decoders, due to the multi-loop approach. Some results unveil that potentially the EL decoder complexity could be reduced, e.g. TMVP was shown to provide less benefit in EL. A good tradeoff between compression performance and complexity should be offered.

More improvement of the complexity measurement procedure is also necessary, some inconsistencies were mentioned during discussions that should be resolved (BoG has been installed).

7.2BoGs

JCTVC-M0436 Report of Non-3V high-level syntax BoG [M. M. Hannuksela]

Initial BoG report provided.

[Add notes as necessary from BoG report.]

• 
  The BoG agreed that the use cases [copy from BoG report]. Agreed.
  
• 
  M0044 is for further study.
  
• 
  M0219 / M0309 / M0274:
  
  • 
    M0219 crops the reference layer. No action.
    
  • 
    M0309 sends offsets (in VPS) to specify the relative spatial alignment of the base & enhancement layers (without cropping the base layer). Regarding "region of interest scalability" – should the EL rectangle include regions without a corresponding base layer position? The prior SVC can do a (somewhat restricted) form of this. The entire BL region can be a subset of the EL region or the entire EL region can be a subset of the BL picture region. Decision: Adopt with that restricted form of "ROI" flexibility, but in SPS syntax rather than VPS (exact location left to editor discretion).
    
  • 
    M0274 proposes to allow inter-layer referencing of the region of the reference layer outside of the conformance cropping window. (The current draft specifies the reference to be bounded by the cropping rectangle.) Decision: Adopted (option A in M0274). Software will be provided for both HEVC and AVC base layer operation by the proponents.
    
• 
  M0180 was basically a subset of M0274.
  

Should scaling factors other than 1.0, 1.5, 2.0 be supported? We currently have this capability in the syntax, but do not have the upsampling filter coefficients for intermediate positions in the draft or SHM.

It would be feasible, but it is not clear whether we need such a profile at this time.

If we don't have the flexibility in a profile, we don't need the coefficients.

It was remarked that we don't need it in the working draft, as it is not clear that we need such a profile.

We currently do not have phase offsets for luma or chroma in the WD or SHM.

Defer to later consideration when it becomes clear whether we need such a profile. Revisit.

JCTVC-M0449 BoG Report on reference layer sample location derivation in SHVC [J. Chen]

For the reference layer sample location derivation used for syntax prediction, the BoG recommended to the method from JCTVC-M0133 with shift bit equal to 16.

For the reference layer sample location derivation used in the resampling process, the BoG recommended the following:

• 
  To adopt the method mentioned in section 2.2 of the BoG report, which is the method used in the current software, as proposed in JCTVC-M0188, JCTVC-M0322 and JCTVC-M0425 (not the entire proposals, but this element of them).
  
• 
  Further study is needed on whether higher accuracy, larger shift bit.
  
• 
  Suggest cleaning up the related software code sections containing the division operation which is disabled by default in the current SHM software.
  

(See additional comments in BoG report.)

Further discussed – chaired by J. Boyce.

Decision: Adopt the solution described above, as recommended by the BoG.
M0455 Complexity

In review of status on Tues a.m., the BoG recommended:

1. 
   To continue AhG17 activity, including the worst case and average complexity evaluation. Need to continue measurement for both memory band-width and computations.
   
2. 
   To report data relative to single layer decoder (on EL resolution).
   
3. 
   In the case of several possible designs within the same basic architecture variation (i.e. different possible designs within pictures-based processing or within block-based operations), it is suggested to assume the optimal (lowest complexity) implementation for the worst case evaluation. (This may include pointer referencing rather than copying.)
   
4. 
   To continue to consider both picture-based and block-based operation (e.g. for upsampling).
   
5. 
   To include memory bandwidth measurement for reconstructed signal writing, MV mapping process and reading-writing during picture-based resampling.
   
6. 
   Use spatial scalability ratio 1.5 for the worst case complexity evaluation (unless it will be decided that SHVC shall support arbitrary scalability ratios).
   
7. 
   Complexity evaluation spreadsheet and s/w module will be released in several versions. Anchor is the same for both but “test” part will contain example of tool in each SCE category. AhG will prepare “anchor” part. SCE coordinators are mandated to take responsibility for “examples” preparation.
   
8. 
   Issues #2 and #3 needs clarification before complexity evaluation methodology can be finalized (for “SNR” case).
   
9. 
   Consider collocated BL picture as an additional reference frame (compare to single layer decoder).
   
10. 
    Interpretation of average complexity data (relation with power consumption) to be studied in AhG.
    

It was noted that the analysis spreadsheet does not address latency issues (e.g. the latency of whole-picture versus block-based upsampling operation).

The BoG recommendations were adopted; further refinement as needed will be done as AHG / CE activity, with the provided spreadsheet for worst-case memory bandwidth calcuation to use as the starting basis.

M0450 Joint BoG on HLS

Ongoing –reported Thurs 25th 1230.

Large BoG report – review notes to be added.


M0460 BoG on perf analysis and non-normative encoding enhancements

Reviewed – notes integrated into contribution-specific sections of the report.