International organisation for standardisation

Requirements meeting report

Source: Rob Koenen, Chair

(Thanks to Bob Bell, Ed Hartley and Mike Zeug for keeping notes)

MPEG-2

PDAM on 4:2:2 @ High Level

It was decided to specify 1152 lines and not 1088, because otherwise the hierarchy with lower profiles would be broken. Having heard the comments from the Japanese and US National Bodies, and recognizing the desire (e.g. by SMPTE and the USNB) to specify 1088 because this is what everybody seems to be using anyway, it was decided to ask National Bodies to comment on the desirability to go to 1088 lines for all high levels.
It was also decided not to take the MB/second approach that is used in MPEG-4. The Requirements group felt that we should stick to the MPEG-2 approach for MPEG-2 Levels, and consider the new approach for all MPEG-4 Levels. All participants agreed that the new approach is useful in principle.

MPEG-2 / MPEG-4

MPEG-2 @ Very High Level & MPEG-4 in the Studio

Higher quality requirements for broadcast, production were discussed, on the basis of input documents (see agenda below for numbers).

It was shown that MPEG-4 syntax can be extended to 4:2:2. There were differing opinions on how MPEG-4 performs at very high level. It was noted that higher quality issues should be separated into levels for Broadcast and for Production/Studio. The latter requires significantly higher quality.

Main Video Requirements were up to 4:4:4 and 10 bit. JNB wants to work VHL in MPEG-4 for simple, main and 4:2:2. USNB wants work on Very High Levels in MPEG-4.

The conclusions were:

• 
  There is a need for a standard addressing digital audiovisual material that cannot be supported by MPEG-2 at the moment. Requirements include higher SNR and/or larger pictures and advanced composition functionality.
  
• 
  It would not be wise to extend both MPEG-2 and MPEG-4 with similar tools or addressing similar applications. This would lead to confusion.
  
• 
  The requirements should be studied further; not just Video requirements but also Systems and perhaps Audio.
  
• 
  After the Requirements Study, MOPEG must decide if, and if yes, when and where (MPEG-2 or -4) to address which requirements. A special AHG was installed for this study.
  

An AHG was installed to deal with the issue. (N2363: Requirement study for High Quality Applications, chaired by Y. Yagasaki and A. Luthra,)

MPEG-4

Visual Profiles and Levels

General

Based on NB comments, other input, and quite some discussion, the following decisions were taken:

1. 
   The change of terminology, from ‘Object Profile’ to ‘Object Type’ and ‘Combination Profile’ to ‘Profile’ was confirmed
   
2. 
   Simple and Core Object types will remain unchanged (on request of several NBs)
   
3. 
   Simple and Core Profiles will also remain unchanged (on request of several NBs)
   
4. 
   It was decided that some tools were OK for addition to the Main Object in principle (1/4 pel, GMC), but the Video Group later on decided they were not ripe, and hence they were not added.
   
5. 
   Levels for the Mesh Object were defined
   
6. 
   Facial Animation was not included in Main; a Hybrid extension of Main foreseen for Version 2, which includes FA. The Requirements Group did not accept the argument that the Face would not have to be rendered, or only barely (a few points). This did however raise the issue of what minimum rendering requirements would be for the Simple Face Object, an issue that could not be resolved during the meeting.
   

Some NBs asked for a clear separation of profiles with natural and synthetic tools. After discussion it was decided not to do this, because there was no consensus to make the change. Reasons for not making the change were also:

1. 
   it would introduce an extra dimension in Profiling, which was seen as undesirable
   
2. 
   for some tools it is unclear where they belong (e.g. is a texture in Natural or Synthetic?)
   

Related to this issue, there was a long discussion about concerns (also from NBs) about the inclusion of the still texture Object in the Main Profile. However, the discussion was not based on evidence but on assumptions. Hence, it was decided not to change Main, and to ask concerned parties to back their concerns with evidence.

New way of defining Levels included in Study on FCD

In a joint meeting with Implementation Studies, the new way of defining levels was discussed. While there was consensus on the desirability of such a new and more flexible way of defining levels, there were concerns for the following reasons:

1. 
   Are the numbers really a good measure in all types of decoders (i.e. Hardware and Software)
   
2. 
   how will Encoders implement the measure?
   

Both the old and the new way of defining Levels is included in the Study on FCD, and a resolution was adopted asking for NB comments:

The Requirements Group recommends that NBs comment on the proposed new way of defining levels using a complexity formula.

Numbers in Levels need cleaning up

will be done in AHG

Audio Profiles and Levels

The Profile and Level situation in Audio did not change much. It was decided to introduce two new Audio Object types:

1. 
   Null: for including local Audio in Scene Graph (= mixing)
   
2. 
   Test: for generating test signals
   

It was confirmed that Parametric Speech should combine the HILN and HVXC tools.

Overview of Profiles and Levels

The Requirements Group agreed that the Profile and level situation in MPEG-4 is hard to follow, and that it would be very useful to have a concise overview that would list all Profiles and Levels defined within MPEG-4. Such a document was drafted by Olaf Barheine (based on his input m3648 MPEG-4 Profiles/Levels Summary) and Kevin O’Connell.