Organisation internationale de normalisation

3AVC base specification, errata, and related topics

The latest prior errata reporting status was provided in the JVT-Y210 output document of the previous meeting. Three additional documents related to errata issues were submitted for consideration at this meeting: JVT-Z025, JVT-Z043, and JVT-Z044.

3.1.1.1.1JVT-Z025 ( Errata 2.0/3.1) [Y.-K. Wang, M. M. Hannuksela (Nokia)] SVC corrigendum items

This document reported three possible corrigendum items for the SVC specification.
The first item is related to a semantics constraint on sub-picture scalable layer SEI message. A fix is proposed.
JVT Decision: Adopted in spirit – exact phrasing to be determined (esp. relating to possibility of presence in different NAL units).
The second item is related to a constraint on the co-existence of “old” H.264/AVC SEI message and “new” SVC SEI message” in a same SEI NAL unit. A fix is proposed.
JVT Decision: Adopted in spirit – exact phrasing to be determined (esp. relating to changing an "and" to "that is").
The third one is on the definition of “decoded picture”. It was proposed to add a definition of “decoded picture” in Annex G to explicitly exclude a “reference base picture” being a “decoded picture”. Remark: Perhaps adding a NOTE (e.g. near semantics of use_ref_base_pic_flag) would be better than adding a new definition.
JVT Decision: Adopted in spirit – exact phrasing to be determined.

3.1.1.1.2JVT-Z043 ( Errata) [H. Schwarz (HHI)] SVC errata

This document contained an SVC related errata list for eventual incorporation into a future amendment or corrigendum to the ITU-T Rec. H.264 | ISO/IEC 14496-10 Advanced Video Coding standard.
This document reportedly started with the JVT output document JVT-X201 as its basis. Changes were relative to that document. The document reportedly included all necessary issues of which the editors were aware prior to the 26th meeting.
Most issues were reported to have been identified by Danny Hong of Vidyo.
"r1" version also provided and presented.
JVT Disposition: Approved ("r1").

3.1.1.1.3JVT-Z044-L (Late Errata 2.0/3.1) [V. Bottreau (Thomson)] On level limits common to scalable profiles – constraint "l"

JVT members supporting presentation:

• 
  S. Pateux
  
• 
  H. Schwarz
  
• 
  S. Gao
  
• 
  Segall
  

Apology? Has been sent.

According to JVT-X201, constraint "l" sets limits to the number of reference layer macroblocks that can be encoded with mbType equal to I_PCM, I_16x16, I_8x8, I_4x4, or I_BL according to the number of enhancement layer macroblocks encoded with mbType equal to I_BL. It is understood that the primary intention of such a constraint was to limit the required decoder complexity. However, it was asserted that the impact of such a constraint may not have been sufficiently investigated. The contribution proposes to discuss the impact of this constraint from an encoder perspective and highlights some use cases that such a constraint reportedly may preclude.
Reportedly, Equation G-370 may preclude some potential use cases. In addition, it was asserted that such a constraint imposes too strict and/or complex encoding rules from an encoder perspective. It was asserted that Equation G-370 sets encoding rules either on the reference layer or the enhancement layer by enforcing either a specific reference layer macroblock pattern to be encoded with mbType equal to I_PCM, I_16x16, I_8x8, I_4x4, or I_BL or an enhancement layer macroblock pattern to be encoded with mbType equal to I_BL. The contribution proposes that:

• 
  Equation G-370 be reformulated in order to better take into account the number of enhancement layer macroblocks eligible to be encoded with mbType equal to I_BL, i.e. for instance only the macroblocks lying within the cropping window;
  
• 
  And/or Equation G-370 be reformulated in order to minimize its impact on encoding mode selection for reference layer macroblocks, for instance by changing the 1.5 arbitrary factor;
  
• 
  Or that constraint "l" be removed.
  

JVT decision: This appears to be a valid report of an actual problem in the standard. The intent was to establish a constraint that constrains (only) the macroblocks in the base layer that are actually used in the decoding process of the enhancement layer. Under some conditions (e.g. involving cropping) the text does not seem to express that intended constraint. Further study may be needed to draft the final necessary correction.