Organisation internationale de normalisation

18Plenary Discussions and BoG Reports

18.1.1.1.1.1.1.1.1Entropy coding dependency discussion

A discussion of inter-frame dependency for parsing was held on Monday 24th at 2 p.m. A chair raised the following question for discussion: If a dependency of entropy decoding from a previous-frame slice were allowed, could it potentially improve the compression efficiency? If it is allowed in one case of motion vector prediction, why not in other cases? Comments made in response included the following:

• 
  It would only be worthwhile if it provides sufficient gain in terms of compression performance
  
• 
  Would need to be switchable in order to operate in lossy environments or for random access
  
• 
  For several slices within one frame, no dependency should exist for the purpose of parallelism
  

Currently this is purely hypothetical topic, but several experts think it could be interesting to perform a study of cases in which this question arises, either in an AHG or CE.

18.1.1.1.1.1.1.1.2JCTVC-D445 BoG Draft of proposed new CE10 on Core Transforms [Pankaj Topiwala, Robert Cohen, Madhukar Budagavi, Rajan Joshi] (BoG draft of proposed new CE registered Monday 24th after start of meeting, uploaded Monday 24th, fifth day of meeting)

This BoG report presented a draft CE on transform design. Remarks made during its discussion included the following:

• 
  QP range should be as wide as possible (down to 0/1 if entropy coder allows)
  
• 
  Test the dynamic range at different points (input/output of quantization/transform)
  
• 
  Regarding "Implementation in software and hardware" – how will the latter be done? Some proponents may not be familiar with the appropriate hardware development/simulation tools – needs further discussion. Also, the period until the next meeting is rather short.
  

18.1.1.1.1.1.1.1.3JCTVC-D443 BoG report on Intra Prediction Improvements [Ali Tabatabai] (BoG report registered Monday 24th after start of meeting, uploaded Monday 24th, fifth day of meeting)

This BoG report concerned intra prediction, and the following comments were made during its discussion:

• 
  Plan to work on possible improvements based on feedback of meeting within next 2-3 weeks
  
• 
  Start work on collaboration bringing several tools together afterwards
  
• 
  Only focus on intra prediction elements (no combinations with elements that would fit in other CEs)
  
• 
  Analysis of complexity: Provide a template to fill number of operations at block level, pixel level, recursions etc.
  
• 
  Separate analysis of luma and chroma
  
• 
  Sequence of reporting concatenation of improvements should be A-bugfixes B-improved signalization C-new modes D- complexity reductions
  
• 
  In case of combinations (e.g. prediction and transforms) results should be reported such that also the standalone performance of prediction can be judged
  
• 
  Need to clarify which version of software to use for CEs
  

18.1.1.1.1.1.1.1.4JCTVC-D441 BoG report of CE9: Motion Vector Coding [T. K. Tan (NTT DOCOMO), W.-J. Han (Samsung), B. Bross (Fraunhofer HHI), J. Jung (Orange Labs), K. McCann (Samsung), Y. Suzuki (NTT DOCOMO), G. Clare (Orange Labs), H. Schwarz (Fraunhofer HHI), A. Fujibayashi (NTT DOCOMO)] (BoG report registered Sunday 23rd after start of meeting, uploaded Monday 24th, fifth day of meeting)

This BoG recommended to adopt PU-level merging with AMVP skip on CU-level (3.1.t) with the following Partial Merging restrictions when a CU has a size greater than the minimum CU size (currently 8x8) is split into 2NxN or Nx2N PUs (JCTVC-D233):

• 
  The first PU is inferred to be merged
  
• 
  The second PU signaling follows 3.1.t (allows either merge or MV coding)
  

Decision: Adopted.

Due to the fact that 3.1.e and 3.1.t are very similar, it was recommended to create a CE on the MV derivation process. The difference between 3.1.e and 3.1.t is that in 3.1.t, skip is enabled with derivation of the MV from MV Coding, in 3.1.e, there is no skip mode but there is a merge with no residual. The adopted design, and the reference of the CE should be:

• 
  3.1.t (Direct off included, which was agreed)
  
•