&%PAGE&

-  &%PAGE&  -
Document AVC-257R
CCITT SGXV                                            	Document AVC-257R
Working Party XV/1                        			March 27, 1992
Experts Group for ATM Video Coding				

SOURCE : CHAIRMAN
TITLE  : REPORT OF THE SIXTH MEETING OF THE EXPERTS GROUP FOR
         ATM VIDEO CODING IN STOCKHOLM/HAIFA (March 18-27, 1992)
         - PART II
Purpose: Report

-----------------

PART II  JOINT SESSIONS


1. General

Joint sessions with ISO/IEC JTC1/SC29/WG11 (MPEG) were held in Haifa at the 
kind invitation of Zoran Microelectronics Ltd during 23-27 March 1992.  At 
the opening session, Dr. Aharon Gill welcomed the CCITT Experts Group.  The 
Experts Group appreciated the support and hospitality of the hosting 
organization.

A list of documents considered during the joint sessions is attached to this 
report as Annex 1.


2. Requirements sub-group (by S. Okubo)

2.1 General

The Requirements Group met on Monday (16:30-18:00) and Tuesday (9:00-10:15, 
14:00-15:00) to discuss requirements for MPEG-2 video, audio and system.  It 
had a joint session with Audio group on Wednesday (11:00-12:30) for planning 
the phase 2 work.  It also had an editorial group meeting on Thursday (9:30-
12:00) for completing the two documents concerning the requirement listing.

2.2 Requirements listing for video

2.2.1 Ad hoc group work

Outcome of the ad hoc group on this subject is contained in MPEG92/130 
(AVC-253) which describes general requirements in Annex 1 and a guide for 
the video work in Annex 2.  Application profile work initiated at the 
Singapore meeting should be continued for various cases because Annex 1 to 
Annex VI (Report of Requirement Group) only shows examples for three 
typical applications.  Ranking of each requirement according to the degree 
of necessity should be used as described in Section 4.2 of MPEG92/130 (AVC-
253).

During the discussion, one particular requirement was pointed out that the 
3:2 pull-down system needs indication of exact field repetition when editing 
is involved.  This fact should be taken into account in the video coding 
design; MPEG-1 does not have this capability.

As to the guide for video work in Annex 2, the meeting recognized the 
necessity of such a document, but agreed to the view that it should be 
shorter, crispening and of a pointing nature. 

2.2.2 Broadcasting applications

The last section of Document MPEG92/141 listed some considerations to 
increase the utility of MPEG-2 for use in broadcast digital video, stressing 
that MPEG-2 should strive for maximum interoperability with broadcast 
digital video, including perhaps HDTV.

Document MPEG92/131 provided an extensive list of user requirements for 
broadcasting purposes and requested MPEG to keep close contact with CCIR.  
The meeting appreciated this input for inclusion in the above mentioned 
"requirements listing" document.

During the discussion, it was clarified that though coding of CG (Computer 
Graphics) may be under responsibility of another ISO group, we should be 
able to sufficiently encode such source materials which include CG pictures 
as part of them.  It was also clarified in response to a question that a 
possible yardstick to the "drastic loss of coding efficiency" mentioned 
related to scalability and compatibility (Item 6 in Section 3/Appendix of 
MPEG/92131) is one point in the five grade subjective scale.

There was some discussion on the broadcasting chain consisting of source, 
studio, emission (or primary/secondary distribution), receiver and recorder.  
A particular concern expressed was the receiver and recorder part.  The 
question was whether received digital signals be recorded directly and used 
for trick mode, editing etc.  One view was that the encoding should be ready 
for such uses.  Another view was that in the short term the received digital 
signals be decoded and analog signals be recorded in the conventional 
equipment and processed as is widely done now, because such digital 
recorders may be equipped with analog input in any case.  This matter be 
further considered, awaiting contributions.

2.2.3 Scalability

Documents MPEG92/141,174,198 and another oral presentation expressed a US 
view stressing the need of scalability feature in the MPEG-2 standard.  The 
meeting renewed recognition of such requirements.

2.2.4 Refinements of the requirements listing

A small group, consisting of A.Lippman, S.Okubo, T.Homma, T.Tanaka, M.Tinker, 
N.Wells, refined the general requirements listing and guide for the video 
work incorporating contributions and views obtained at the meeting at 
large. 

The outcome has been issued as two separate documents; MPEG92/229 and 230.

2.3 Materializing the "generic standard" philosophy through the TM work

At the Singapore meeting, it was agreed that the objective of our work 
should be to achieve the "single standard" as far as possible, but in 
practice the "toolkit with maximum core" may be the solution.  There was 
also raised a notion of "supercore" for consideration.

The question for this meeting was how to reflect this objective into the 
Test Model work currently being carried out.

MPEG92/154 (ISO/IEC Technical Report) provides a concept of "profile" for 
possible use in our standard, which defines combination of "base standards" 
and particular subsets of options to provide specific functions.  MPEG92/188 
(Annex 5 to AVC-256R) proposed provision of a test method to optimize the 
low delay mode in the Test Model experiments.

Taking these input contributions into consideration, the meeting reached a 
common view as follows;

1)	Since the generic standard is intended to meet the performance and 
functionality requirements of each particular application, one way is to 
emulate the situation in the Test Model work.

2)	Each contributor may have some application in mind as user of the 
standard and be willing to do his/her best efforts to optimize the Test 
Model for the particular application, Test Model refinement should be 
integration of such efforts in a harmonized way.

3)	To avoid divergence in the current convergence phase work, above 
mentioned improvement efforts should be done using a single syntax defined 
in the Test Model.

4)	Based on the analysis to relate functionality with necessary elements 
in the standard, the following items are suggested for inclusion in the Test 
Model experiments;

   a. Basic performance
   b. Scalability
   c. Low delay
   d. Simplest decoding (for such as broadcasting environments)
   e. Robustness to bit error and cell/packet loss
   f. MPEG-1/H.261 compatibility
   g. Trick mode
   h. Extension to higher formats

5)	The improved syntax should be reviewed each time to implement the 
"maximum core" objective as well as to reflect the coding efficiency and 
implementation consideration.  Use of good engineering sense is expected.

2.4 Procedures to improve the Test Model

A process from a new idea to its inclusion in the Test Model was discussed 
and clarified as in the separate document WG11 N0158 (see Annex 2 to this 
report), which defines a process from a new idea to its inclusion in the Test 
Model through experiments, documentation, core experiments and 
verification.

2.5 General requirements for audio and system

2.5.1 Purpose

This part of Requirement session was the first free discussion occasion to 
obtain an overall view of the MPEG phase 2 as a total system.  Main topic 
was identification of general requirements which can not be met by the 
MPEG-1 standard.

2.5.2 Alignment of the time schedule

At the moment, discussion is directed toward aligning audio and system works 
with that of video; freezing the specifications at the end of 1992 and 
official standard in 1994.

2.5.3 Necessary works for MPEG phase 2

One area identified as needing additional work is "multi-channel" system 
which may be a video with several high quality audio channels and 
associated multiple language channels, or furthermore a set of multiple such 
programs.

CCIR study programs for such multi-channel audio systems were brought to 
attention of the meeting.

Audio coding experts expressed that a possibility to cope with this need 
under the phase 2 time frame is to extend the current standard, by 
exploiting the characteristics of multiple coded sources as well as by 
lowering the sampling frequency for e.g. language channels.

Another item raised for consideration is constraints in the satellite 
broadcasting; noisy channel, switching bitstreams, adaptive multiplexing, etc.

For the system part, there were two input documents for this meeting which 
proposes to work on error resilience (MPEG92/088) and control & indication 
signals (MPEG92/089). 

It was also pointed out that during the video requirement survey, the 
following items had been identified as additional requirements to the system 
(see Section 4 of Annex VI to the Singapore meeting report);

    - signal encryption and/or scrambling for authorized receivers etc.
    - mechanism to implement backward compatibility between systems with 
      different bit rates 
    - editing coded signals in such applications as home video 

Since this discussion was of a general nature, deeper considerations were 
expected in the individual audio and system sessions.

2.6 Work plan for Audio

Joint Audio/Requirements meeting reviewed the possible work items taking 
into account the time frame set for overall MPEG phase 2.  After some 
discussion, the following study items were agreed upon as targets achievable 
by means of extending the MPEG-1 audio coding;

    - compatible commentary grade audio coding (Note) 
    - compatible multi-channel audio coding (multi stereos, multi languages; 
      Note)
    - transcodability

		Note: These may allow scalability.

Furthermore, the following items were identified as requiring longer term 
study;

    - lossless coding
    - 7kHz videoconferencing
    - fast mode
    - ATM mode
    - N-ISDN distribution;

See the Report of Audio group for more details.


3. Video sub-group (by G. Bjoentegaard)

3.1 General

The activity in the video group was divided into three parts:

    -	Comments to the MPEG-1 CD.
    -	Conformance testing.
    -	Work towards TM1.

The first two points are not considered to be of specific interest to the 
CCITT group.  The focus will therefore be on the third item which reflects 
the actual collaboration between CCITT and MPEG.

3.2 Work towards TM1

A preliminary document (PWD) for definition of the first test model was 
available prior to the meeting.  The goal of the HAIFA meeting was to 
complete the definition of the first test model (TM1) and define experiments 
to be performed until the next meeting.

3.3 Procedure for improvement of the test model

The procedure for incorporating new ideas in the test model was clarified 
(see Annex 2 to this report):

    -	Assume that a promising idea is presented at a meeting.  If more than 
one lab finds the idea interesting for further testing, sufficient 
documentation is included in the test model to be able to make 
comparative tests at different labs.

    -	If the result of this test is positive (the definition of "positive" is 
not clearly defined) the idea is incorporated in the test model.

According to the present time schedule, it will be very difficult to alter 
the TM significantly after November 1992.  This means that ideas produced 
after July 1992 will have a minimal chance of getting into the test model.

3.4 Profiles

The coding standard shall be "generic".  It must therefore fulfil 
requirements set by different applications.  In order to ensure that all the 
requirements are fulfilled, a set of profiles is defined.

A profile is a subset of the features defined in the test model and is 
intended to correspond to specific applications.  The intention is that  
experiments shall be performed within each profile to ensure that the 
standard fulfils the corresponding requirements.  

Four different profiles have been defined:

    -	High quality profile.  Within this profile there is no constraint 
concerning e.g. scalability, low delay etc.

    -	Compatibility.  The main restriction within this profile is 
compatibility with MPEG-1 and/or H.261.  This profile is also somehow 
connected to spatial scalability.

    -	Scalability.  Within this profile the constraint is more general 
scalability. 

    -	Low delay.  The main applications connected to this profile is two-way 
communication.

Definitions of the different profiles and corresponding core experiments 
were produced at the meeting and will be part of the revised version of the 
TM1 description.

3.5 Prediction modes

Many experiments had been performed and the results were presented at the 
meeting.  Most of the experiments focused on improving the picture quality.  
Much focus was on the use of different prediction modes.  A small group was 
therefore set up to define the set of prediction modes to be included in 
TM1.  This resulted in a considerable increase of prediction modes and 
therefore also complexity of the test model.  The intention is that 
experiments shall make it possible to point out the most useful prediction 
modes and that a large amount of the others may be discarded at a later 
stage.

3.6 Ad hoc groups 

Several ad hoc groups were formed for doing collaborative work until the 
next meeting.  The groups of most interest for CCITT are:

    -	TM1 editorial group. 	(Arian Koster)
    -	Compatibility.          	(A. Koster)
    -	Scalability.            	(C. Gonzales)
    -	Quantisation.          	(N Wells)

3.7 Items of particular interest to CCITT

Of the four profiles, the ones on compatibility (with H.261) and low delay 
seems to be of particular interest to CCITT.  In addition it is important to 
focus on cell loss characteristics in connection with ATM.  

The new prediction modes is of particular importance for the low delay 
profile.  The increased number of choices will give better predictions also 
for P-frames.

	Authors note: In order to get good low delay prediction modes, CCITT 
members must be very active both in presenting simulation results and 
in the meeting.


4. System sub-group (by B.G. Haskell)

The MPEG Systems Committee spent most of its time reviewing comments on the 
MPEG-1 Committee Draft.  Subjects covered briefly during a short meeting on 
requirements included error correction, cell loss, and encryption.


5. Implementation Study sub-group (by D.G. Morrison)

5.1 Sessions and attendance

                    25th am    26th pm    27th am
John Gooding           x          x          x
Colin Smith            x          x          x
Kwok Chau              x          x          x
C J Lee                x          x          x
Marco Gandini          x          x          x
Sam Narashiman         x          x
Ben Yung               x
Eiichi Kowashi                    x
Don Shaver                        x          x
Geoff Morrison         x          x          x

5.2  Comments on TM0

Many participants had not seen the preliminary test model agreed at the 
Singapore meeting.  Clarification of several aspects of 'field' and 'frame' 
motion compensation and coding were sought but could not be adequately 
answered from the available documents.  The meeting decided to join the 
Video subgroup meeting on Wednesday afternoon where presentations on 
these aspects were taking place.  The Implementation Studies subgroup
subsequently were of the opinion that this area was becoming rather complex 
and drafted some resolutions to voice this concern and suggest joint ways 
forward with the Video subgroup. 

Several organisations had found a problem with chrominance vertical sample 
rate reduction and six solutions had been proposed in the Video subgroup.  
The Implementation Studies subgroup should look at this topic to see if 
there is a preferred solution from the complexity viewpoint.

5.3 Implications of scalability and compatibility on implementations

The video subgroup was receiving input documents on scalability and 
compatibility and some of these were drawing conclusions about the effect 
on implementation complexity.  These should be checked by the 
Implementation Studies subgroup.  As these issues could not be adequately 
considered during the meeting they were deferred.  Topics to be covered 
include:

    -	frequency domain separations
    -	pel domain separations
    -	separation of coded bits performed by video multiplex or system 
multiplex.

5.4 Others

1)	The subgroup decided to attend the meeting on compliance on Thursday 
morning (26th).

2)	It was recognised that lack of a quantitative methodology for 
assessing implementation complexity is a hindrance to the work of the 
subgroup.  For the benefit of newcomers, the procedures used to assess 
proposals in the competitive phases of MPEG-1 and MPEG-2 were related and 
their strengths and weaknesses discussed.  Though the difficulties were 
acknowledged there was interest in continuing to find better methods.

3)	It was agreed to seek the establishment of an ad hoc group to 
continue working on the above issues until the July 1992 meeting.

4)	Recommendations were prepared for the WG11 plenary to be held in the 
afternoon of 27th April.  These are reproduced elsewhere, incorporated with 
all those from WG11. 


END