&%PAGE&

-  &%PAGE&  -
Document AVC-106R
CCITT SGXV                                            	Document AVC-106R
Working Party XV/1                        			August 23, 1991
Experts Group for ATM Video Coding

SOURCE : CHAIRMAN
TITLE  : REPORT OF THE SECOND MEETING OF THE EXPERTS GROUP FOR
 	    ATM VIDEO CODING IN SANTA CLARA (August 14-23, 1991) 
	    - PART I 
Purpose: Report

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

PART I - INDEPENDENT SESSIONS

				1. General
				2. Documentation
				3. Tape demonstration
				4. Discussion
				5. Work plan and work method
				6. Actions to other groups
				7. Others


1.	General

The third meeting of the Experts Group was held in Santa Clara (USA) at the 
kind invitation of AT&T, Bellcore, C-Cube Microsystems and CLI.  The first 
part of the meeting, held during August 14-16, consisted of CCITT 
independent sessions, while the second part, held during August 14-23, 
consisted of joint sessions with ISO/IEC JTC1/SC2/WG11 (MPEG) and a closing 
session.  The list of participants appears at the end of this report.  

At the start of the meeting, Mr. Haskell made a welcome address on behalf of 
the hosting organizations.  

At the closing session, Chairman thanked the hosting organizations for the 
meeting facilities provided and the excellent secretarial support.


2.	Documentation (TD2)

For this meeting, 41 AVC-numbered documents and 11 temporary documents 
were available as listed in Annex 1.


3.	Tape demonstration (TD4)

Several video tape demonstrations were given with D1 or U-matic to present 
experimental results as detailed in Annex 2.


4.	Discussion

4.1 Reports of the meetings relevant to the Experts Group

	4.1.1 Paris meeting of the Experts Group (AVC-65R, AVC-66)


The meeting reviewed the report of the closing session and the joint 
sessions at the Paris meeting.  As to the discussion of the joint work 
method, see ��6.3 of this report.  Mr. Sandgrind called attention to Mr. 
Bjoentegaard's comment on the time schedule of the Santa Clara meeting of 
Video sub-group and necessary reduction of number of proposals to the 
Kurihama test.

	4.1.2 CMTT/2 Special Rapporteur's meeting (AVC-67)

Mr. Carbiere gave on overview of the liaison statement which has been sent 
to CCITT SGXV and ISO/IEC JCT1/SC2/WG11.  Annex of AVC-67 was consulted 
during the discussion of requirements to the H.26X codec.    

Chairman's Note - During the review of AVC-67, there was a question whether 
486 active lines per frame for CTV (525 line system) is correct or should 
read 480.  According to CCIR Rep.624-3, blanking period for System M is 
defined as 19 to 21 lines which lead to 242.5+/-1 active lines per field.  Rec. 
723 source coding covers 248 lines per field.

	4.1.3 SGXVIII meeting (AVC-68)

This liaison statement to this group contain study results on CLP issues, 
UPC issues and IVS baseline document.  Each of which was discussed at the 
corresponding discussion item below.

4.2 Picture format

	4.2.1 Number of formats (AVC-102,82; AVC-73,86; TD-4) 

Mr. May clarified that "other formats" in "A future standard should leave 
the door open for use of other formats" mean such larger formats for 16:9 
aspect ratio signals and HDTV signals.  

Doc. AVC-82 provided pros and cons for interlaced and progressive formats.  

Mr. Stevens clarified the Australian proposal in AVC-36 as well as AVC-73 
that a maximum sized container is defined for each spatio-temporal 
resolution class, any formats are used as far as they are inside the 
container, and necessary conversions for the local display is carried out at 
the decoder.  These conversions may be covered by the multi-scan capability 
or non-refreshing nature of the display.  There was a question whether side 
information is transmitted to indicate the actual picture size in and mapping 
to the container or black bordered pictures filling the container are 
transmitted.  There was also expressed concern about the difficulty of 
negotiation of the actual picture format size.  Another question was whether 
this maximum container be 16:9 aspect ratio.

	4.2.2 Conversion between CCIR601 and SCIF (AVC-80)

A work was presented to verify the feasibility of format conversion between 
525 line CCIR601 signals and SCIF signals as in AVC-29, where converted 
pictures are evaluated in SNR as well as observation on interlace monitors.  
The following comments were given;

    -	Care should be paid to SNR evaluation, since simple line repetition 
may provide infinite value for interlace-progressive-interlace 
conversion.  observing pictures are most important.

    -	If a progressive picture is made from a field, black and white stripe 
of maximum vertical resolution produces all white and all black 
alternating pictures. 

    -	Observing on a progressive scan monitor is a method. 

    -	If progressive scan signals are observed on interlaced monitors, 
moving slanted lines look stairstepped.

    -	A similar evaluation is ongoing for the 528 lines x 704 pels square 
pixel SCIF in Bellcore. 

The meeting recognized that further study is necessary on the following 
combinations for the SCIF with parameters 576 lines, 59.94 frame/s, 
progressive;

      Source		     Transmission		      Display

	576/50 1:1 ----> |    |       |	  | ----> 576/50 1:1
	576/50 2:1 ----> |    |		|	  | ----> 576/50 2:1
	480/60 1:1 ----> |SCIF| ---->	|SCIF-1| ----> 480/60 1:1
	480/60 2:1 ----> |    |		|	  | ----> 480/60 2:1
	576/60 1:1 ----> |    |		|	  | ----> 576/60 1:1

	4.2.3 Coding efficiency (AVC-79,83)

Two experimental results were presented which address coding loss due to 
the use of SCIF as coding format, one using a RM based algorithm while the 
other using a MPEG1 based algorithm.  Some coding loss has been observed, 
but possibility of its reduction is also indicated.

Further work is requested for better understanding of the problem.

	4.2.4 Progressive scan devices (AVC-81)

Document AVC-81 summarized the present situation for progressive camera 
products in Japan for information.  There was a comment that camera 
availability only at this moment should not be argued for our decision on 
coding format, we should look at coming years.   There was also some 
discussion on the sensitivity of progressive scan camera which is relevant 
to the brightness of displays in videoconferencing environments.
    
	4.2.5 Conclusion (TD5,6)

After having further discussion on the following items;

    - single coding format or multiple ones for higher quality video coding?
    - possible parameter values?  how related to CIF?
    - extension to larger formats such as EDTV and HDTV?
    - service integration aspects
    -	take the initiative to adopt the single format approach (if viable) to 
other groups?

Mr. Haskell undertook to coordinate a small group meeting in the evening of 
August 14 to sort out the problems, to identify  possible solutions as well 
as future action points.  Interim common understanding of the group and 
home work were itemized as in Annex 3, which also contains a table 

elaborated by Mr. Schaphorst showing the framework for discussion of H.26X 
format.

Finally, the meeting agreed to reach an agreement on the picture format 
issue at the earliest occasion in 1992.

4.3 Framework for H.26X

	4.3.1 Requirements

1) Current status (��4/AVC-96, ��3,4/AVC-70)

Chairman reminded the group of the requirements on the H.26X codec which 
had been identified up to this meeting using relevant sections of the 
group's status report (AVC-96) and the Proposal Package Description 
document (AVC-70).

2) Application (AVC-78) 

Australia pointed out "video surveillance" form an application category 
which is not covered by the existing list.  Its impact on the video coding is 
also discussed.  The meeting agreed to add this category in Table 1 of the 
status report and to use this as an input document for the joint sessions 
with MPEG.

3) Secondary distribution (AVC-67)

AVC-67 provided preliminary functional requirements for secondary 
distribution of digital TV and HDTV signals.  It was pointed out during our 
meeting that "channel hopping" is required for receiving television programs 
and that the secondary distribution decoder may require H.261 compatibility 
in the light of common use of a terminal for different services.  These 
points and H.26X requirements as in Annex 5 of this report are sent for 
action in a liaison statement to CMTT/2 Special Rapporteur's Group (see ��6.2 
of this report).   

4) ATM network aspects (AVC-72,84,85,86; AVC-68)

AVC-72 proposed an additional list of requirements, and to include a 
complete list in the IVS Baseline Document.  The meeting studied the content 
item by item, and agreed in principle to the proposal.  Since picture format 
related matters are still under discussion, however, they are excluded for 
the moment.  For the complete list, see 5) below of this report.  

During the discussion, Mr. Verbiest pointed out that the definition of peak 
rate in Annex 2 to AVC-68 becomes ambiguous when NT2 multiplexes cells from 
multiple sources.  Mr. Tabatabai also expressed his question on the network 
performance related to CLP bit as in Annex 2 to AVC-68.  These points are 
sent to SGXVIII as questions (see ��6.1 of this report).   

AVC-84 proposed a specific action to the Kurihama test; submission of 
additional materials to demonstrate cell loss tolerance of proposed 
algorithms.  Some members expressed their concern about additional work 
required before the deadline.  After having considered this factor and 
importance of early consideration of cell loss for the joint development of 
a generic coding scheme, the meeting agreed to propose the modification of 
PPD document as in Annex 4.

During the discussion, it was felt necessary to establish appropriate 
method(s) for evaluating cell loss resilience in differing coding schemes 
(e.g. layered or not), particularly its visibility point of view for short test 
sequences.  Cell loss ratio, distribution, decoding method, etc. should be 
specified.  This matter needs further study of the Experts Group.    

AVC-85 provided a discussion material for setting the coding/decoding delay 
objective to the H.26X codec.  Various comments were given as follows;

    -	The system should not consume all the 400ms allowable end-to-end 
delay.  A much smaller value should be set for the target.

    -	The current "less than about 150ms" target comes from the HRD 
definition (buffer size = 4 times CIF picture period).

    -	Delay is one of the performance items equally important as picture 
quality.

    -	Codec delay is highly dependent on implementation, thus it should be 
left to equipment provider's choice, which contributes to product 
differentiation.

    -	There might be a chance that propagation delay be negotiated as part 
of QOS?

    -	As to the impact of terrestrial and satellite transmission, local 
communication may use higher bit rate terrestrial transmission while 
long distance communications may use low bit rate satellite 
transmission due to the cost reason.  The latter combination worsens 
the total system delay since low bit rate operation of the codec tends 
to induce longer delay.

    -	Delay in AAL is yet to be clarified.

Since the meeting felt it premature to make any decision at this moment, the 
current wording for the target is left as it is, with clarification of the 
background as described in the second hyphenated item above.  At the same 
time, the meeting feels we should strive for smaller delay in H.26X at higher 
bit rates.

AVC-86 and AVC-87 attempted to extract specific requirements on H.26X from 
the service integration and multipoint points of view.  The meeting welcomed 
this input as useful basis for further discussion.  Members are requested to 
contribute to deriving more concrete impacts to the video coding which lead 
to a better architecture.

During the discussion related to the bandwidth flexibility of B-ISDN, it was 
queried how a particular bit rate for each call can be set up since the bit 
rates in B-ISDN are not quantized as in N-ISDN.  Negotiations with both of 
the network and the remote terminal may be involved.  This matter awaits 
further clarification.

5) Conclusion

Based on a number of requirements listed in those contributions, Mr. 
Brusewitz undertook to coordinate a small group meeting in the evening of 
August 15 with the task of summarizing these requirements on H.26X codec 
and H.32X terminal.  The outcome is contained in Annex 5. 

	4.3.2 Architecture

1) ATM specific boundary conditions (AVC-71) 

AVC-71 raised the question of whether source coding and transmission coding 
be separable or not in ATM environments.  During the discussion, the 
following ATM specific considerations were pointed out;

    -	Source shaping by rate control is necessary in response to the UPC.
    -	Use of CLP impacts the source coding.  If the cell loss rate is very 
low, simple error resilience measure in the source coding is sufficient.  
If it is very high, serious preventive measures are necessary in the 
source coding.  Otherwise, cell loss can be prevented in the channel 
coding.
    -	Efforts to reduce redundancy in the source coding can be traded off 
with efforts of network utilization such as the use of different 
priority channels.
    -	Once the separation strategy is taken at the start, later inclusion of 
preventive measures at the source coding makes the total performance 
suboptimal if it be found necessary. 

However, if cell loss ratio can be very low, and if the codecs for STM are 
considered as having VBR properties in number of bits per frame being 
changeable, another scenario may be also drawn where a traditional 
architecture is applied even in ATM environments.  Considering that the 
network characteristics are not fully known, the meeting agreed to proceed 
having both strategies in scope for the moment.

2) Flexible layering (AVC-73,74; AVC-86; AVC-94,100,103)

AVC-73 gave further thoughts on the "flexible layering" introduced in AVC-
35 at the Paris meeting, stressing the identification of service class 
boundaries.  AVC-74 provided a supporting evidence.  Interworking among 
different service classes are realized by switching off constituent layers as 
necessary. 

There were some comments on the experimental data;

    -	Some concerns were expressed concerning layer 1 prediction loop drift 
(between encoder and decoder) under the influence of low priority 
channel cell loss.  In this implementation (AVC-74), method of 
controlling the impact of this drift effect must be studied.
    -	Rate-distortion curve should better be evaluated using a two-
dimensional VLC for DCT coefficients.  Significant difference in INTRA 
mode between the single layer and two layer schemes needs 
clarification.

As general comments, it was pointed out that coding efficiency and hardware 
complexity be carefully studied and that mixture of frame rates in the two 
layers (e.g. 30 Hz of H.261 in the first layer, 25 Hz of CCIR-601 in the second 
layer) may affect coding efficiency.   

4.4 VBR vs CBR 

	4.4.1 Characteristics 


1) Delay estimation (AVC-90)

AVC-89 provided a simplified model to analyze the delay in VBR modes.  From 
comparison with AVC-56 presented at the Paris meeting, Mr. Haskell commented 
that this is an alternative approach to AVC-56, that implications are high 
speed operation and large capacity of memory in the decoder and that coder 
operation time "T" can be smaller in practice.

It was also pointed out that if the window size of network Swin is small, 
coding delay becomes large because of necessary smoothing of peaks in the 
terminal, that the leaky bucket will produce similar results.

2) Required window size for average rate monitoring in the network (AVC-89)

Experimental results were presented for a particular rate control system, 
indicating the window size of several seconds can improve VBR mode picture 
quality.  

After some questions and answers on such items as visibility of coding-
stopped macroblocks, the meeting recalled the importance of the window size 
of the network for average monitoring.  The decision of this parameter 
should be based on interaction between this group and SGXVIII, perhaps with 
iteration.  Mr. Verbiest stated that the Belgian experimental system uses a 
more sophisticated policing with multiple leaky buckets where not only peak 
and 100 sec average rates but also higher order moments are also monitored.   

	4.4.2 Statistics (AVC-99)

Mr. Verbiest offered measured statistics on VBR operation which are stored 
in a floppy disk.  This can be used freely as far as the purpose is research 
and reference is given to RTT Belgium. 

It was recognized that similar data are missing when motion compensation is 
included.  Members are encouraged to provide such data.

	4.4.3 Network loading model (AVC-75,97; Annex 5 to AVC-96)

Improvements to the first network model were discussed.  As pointed out at 
the Paris meeting (see ��4.4.3/AVC-65R), the current model provides 
saturation probability Psat.  The meeting agreed to use the following 
equation for calculation of cell loss ratio CLR;
           
	CLR =   L(R)*p(R)dR /   R*p(R)dR

where L(R) is loss function that can be approximated as Eq.(1) in AVC-75.

The meeting also agreed to modify the network loading model as in Annex 5 
to AVC-96, assuming that its CLR is derived according to the above equation.

There were some discussion on the relation between this network modeling 
and the mechanism according to which the network accommodates calls.  AVC-
97 argued that its model is reflecting a likely mechanism in the network, but 
possibility of using peak, mean and variance was also expressed by others.

It was also discussed that call acceptance mechanism and exact modeling of 
the source and network should be separated, important thing be to obtain 
better understanding and deeper insight to the statistical multiplexing 
problem.  The meeting agreed to this view by noting in the network loading 
model description that we need more accurate model(s) for the precise study 
of VBR vs CBR issue.  Suitable models for this purpose need continued study.

	4.4.4 Guidelines for further work

The meeting confirmed that the guidelines established at the Paris meeting 
for VBR vs CBR issue still stands.  Further study according to these 
guidelines are awaited, particularly experimental evidence for VBR 
advantage in reconstructed pictures under UPC.

4.5 Video coding algorithms

4.5.1 Layered coding (AVC-94,100,103)

Three layered coding simulation results were presented, all of which can 
work in non-compatible mode by switching off the base layer.  They do not 
exactly follow the structure in Figure 1/AV-65R in which the second layer 
codes the difference between the first layer coded picture and the original 
input.  The following points were clarified through questions and answers.

AVC-94

    -	Five motion vectors are calculated for a 32x16 macroblock of 601 
resolution, one for the inner loop, four for the outer loop.

AVC-100

    -	Delay due to "upconversion" can be minimal since it need not be 
subjected to encoder and decoder buffer delay.
    -	This structure allows 16:9 aspect ratio in the MPEG-2 coding.

AVC-103

    -	This result shows that there is some coding loss due to the use of two 
layers, but not so big as anticipated.  Whether the layering can 
compete with "simulcast" in coding efficiency needs further study.
    -	Flicker is not observed even if SNR differs by 4 dB between odd and 
even fields.
  
4.5.2 New elements (AVC-88,104)

Two study results were presented for information on coding efficiency of a 
subband coding and improvements on forward prediction.  There was a 
question whether "CBP" is effective also for high resolution video coding. 

4.6 Multimedia multiplexing (AVC-76,91,92)

The meeting adopted the proposal in AVC-76 that we request of SGXVIII to 
develop appropriate signalling and control for bounded cross media delay.

AVC-91 provided a discussion material for realizing H.32X terminal 
interworking with H.320 terminal focusing on multimedia multiplexing methods.  
It is argued that user multiplex be a solution.  Mr. Morrison pointed out 
that not only H.320 terminals but also MPEG standard sources may require 
the same consideration.  Further study is requested.

AVC-96 provided a discussion material on the virtual layer approach to 
define multimedia multiplexing.  The meeting welcomed this useful input and 
requested of the members further development of the idea.  Mr. Stevens 
suggested the naming would be more suitable if it is "media control layer" 
or "media convergence layer" instead of "MUX layer".  

4.7 AAL Type 1 and Type 2

	4.7.1 Current status (AVC-68)

Chairman gave a brief summary of the AAL study in SGXVIII referring to 
Annex 2 to AVC-68.

	4.7.2 Required elements for audiovisual communication systems (AVC-77)

The meeting supported the considerations in AVC-77 that AAL Type 2 be 
commonly used for VBR and CBR audiovisual communications, and recognized 
the urgency of study on the definition of AAL Type 2 by the Experts Group.  
The meeting also agreed to place relevant questions to SGXVIII in the 
liaison statement to SGXVIII (see ��6.1 of this report).

	4.7.3 Error and cell loss protections (AVC-98,69)

AVC-98 provided information on bit error and cell loss performance required 
for video services assuming a certain duration of error free time for each 
service.  Preventive measures are exemplified in AVC-69.  There were some 
concerns whether simple error free time calculation help us because 
visibility of the bit error and cell loss much differs among coding schemes; 
e.g. according to being layered or not. 

Since the preventive measures against bit errors and cell losses have 
crucial impacts to the structure of audiovisual terminals for B-ISDN, the 
meeting agreed to expand the content in relevant parts of the IVS Baseline 
Document (p.20 and p.22 of AVC-68) covering the above concerns by notes as a 
step in the ongoing iterative process.

There was discussion on where necessary error correction and cell loss 
protection be carried out, at SAR, CS or user layer depending on the 
network performance.  The following factors should be considered in this 
study:

    -	Interworking with N-ISDN terminals should also be considered where 
bit errors are taken care of at the user layer.
    -	It should be noted that scrambling in ATM layer causes correlated 
errors in SAR-PDU.
    -	Order of scrambling, error correction and encryption should be 
carefully considered.
    -	If video is layered, different layers may take advantage of different 
levels of bit error and cell loss performance.

4.7.4 Comments on AAL Type 1 (AVC-93)

The meeting agreed to include the requirements to the AAL Type 1 functions 
raised by AVC-93, with necessary modifications, in the liaison statement to 
SGXVIII (see ��6.1 of this report).

There was a question whether audiovisual terminals use Type 1 as well if 
both VBR and CBR are covered by Type 2 as in AVC-77.  One of the cases 
identified was interworking between B-ISDN and N-ISDN, which was discussed 
in more details (AVC-91, TD-7).  The meeting felt ambiguous about whether 
Type 1 for interworking be required in H.32X terminal, or it may be covered 
by the interworking unit in the network.  It was also pointed out that the 
existing N-ISDN terminals are quite susceptible to cell loss if it occurs in 
the circuit emulation mode.  These points need advice of SGXVIII.

4.8 IVS Baseline Document (AVC-68,101; AVC-72,98)

It was agreed to adopt the proposal in AVC-101 as input for upgrading the 
IVS Baseline Document, which are sent to SGXVIII with other points identified 
during the meeting (see ��6.1 of this report).


5.	Work plan and work method 

5.1 Harmonization with other groups (AVC-95)

The meeting had opportunity to have attendance of Mr. Yasuda who is acting 
as Convenor of ISO/IEC JTC1/SC2/WG8.  He presented the current organization 
under his responsibility, focusing on the recent establishment of new SC29 
where all the audiovisual and hypermedia coding groups are assembled, and 
requested continuous cooperation between these ISO groups and relevant 
CCITT groups.

5.2 Collaboration work with MPEG

The meeting had a free discussion session on possible improvements of the 
joint work method with MPEG.  This session was continuation of the relevant 
part of the Paris meeting closing session as recorded in ��6.3/Part I of AVC-
65R.

The following voices were heard on the current situation as needing 
improvements;

    -	MPEG is still spending much time for MPEG-1.

    -	From practical point of view, CCITT sole sessions spend 3 days without 
sufficient time for discussion, while joint sessions continue 5 days for 
topics of our less concern.  This lacks balance.  Closing session 
should be held at an earlier occasion so that attendance to the joint 
sessions become more flexible.

    -	MPEG meetings are too frequent, thus close coupling of our meetings 
with them may cause travel budget problems in some organizations, and 
furthermore shorter interval between meetings may decrease high 
quality contributions.

    -	Meeting place of sole sessions should be more carefully chosen.  For 
example, this meeting in Santa Clara could have been in the East coast 
where better support of the meeting becomes available.  It is not 
always necessary to meet at the same place with that of joint 
sessions.

    -	Document managements need improvement.

    -	Use of liaison representatives may solve these problems.  (Chairman 
noted that appointed representatives to respective sub-groups are 
not meant to this "liaison representatives" but should function as co-
chair with that of MPEG.)

    -	That method of collaboration may end up with MPEG's sole 
responsibility of joint development of the generic coding.  
Furthermore, CMTT which adopts this way of work is less influential to 
MPEG.  It is noted that broadcasters have become more influential in 
MPEG-2 through more participants from this field compared to MPEG-1.

    -	Agreements obtained through discussion are not clear enough.  
Decision process looks different among sub-groups, depending on the 
character of the person in charge.  

    -	Change of terms of reference needed more time for discussion because 
the item was new to recently joined CCITT members.
  
    -	Due to the large number of participants, single persons have 
difficulty to make his voice heard, and to identify how decisions were 
made.

    -	We should start from knowing each other, we may need some time to 
achieve this, working together toward a common objective may increase 
this possibility.  Sharing a common Test Model for collaborative work 
plays an essential role for this purpose.  Enhanced presence of the 
CCITT Experts Group should be sought through good technical works. 

    -	Having a technical session on communication aspects such as ATM may 
help mutual understanding among members ready for joint development.

    -	Efforts of penetrating the joint development spirit through the 
management may help us.  Mr. Yasuda undertook to reflect this point 
as part of organizing the new SC29.

    -	One difference in attitude between ISO and CCITT is that ISO is more 
oriented to the definition of bit stream while CCITT also addresses 
terminal and overall service matters.  This difference may be causing 
some mismatches between their approaches.

The meeting confirmed to review this matter on a continuous basis, taking 
into account the facts that we will experience our sole sessions after the 
joint sessions at the next meeting in November and that the Kurihama 
sessions will involve presentation of proposed algorithms which are of our 
great concern.  

At the closing session on August 22, Chairman presented Mr. Chiariglione's 
letter to him in response to these voices which had been raised at the MPEG 
Chairmen meeting on August 18, containing record of discussion at the MPEG 
Chairmen meeting and some other comments.  The meeting welcomed this 
message for our mutual understanding. 

Chairman concluded this discussion by confirming that we will continue our 
efforts toward the success of joint work through grass roots activities as 
well as top-down actions.

5.3 Collaboration work with CMTT/2 SRG

The Experts Group invites increased level of participation of CMTT/2 SRG at 
all meeting in a liaison statement (see ��6.2 below of this report).


6.	Actions to other groups

6.1 SGXVIII (TD-9)

Mr. Tanaka drafted a liaison statement containing the following items found 
during the meeting;

    - further questions on the network characteristics
    - request for multimedia support
    - requirements on the network characteristics from the video coding 
    - comments on AAL Type 1 and Type 2
    - improvements of the IVS baseline document

The meeting approved this draft with some amendments as in Annex 6.

6.2 CMTT/2 (TD-10)

Two liaison statements are sent to CMTT/2 SRG for action as in Annex 7; one 
contains comments on their preliminary functional requirements, the other 
contains invitation to the joint work.

6.3 MPEG 

1)	The list of H.26X requirements (Annex 5) and the proposal of additional 
submission material concerning cell loss (Annex 4) are input to the joint 
sessions.

2)	The meeting appointed representatives to MPEG subgroups (Video, 
System, Requirements, Test, Implementation) as follows;

	MPEG-VIDEO (Didier LeGall)			J. Guichard
MPEG-SYSTEM (Sandy MacInnis)			B.G. Haskell
MPEG-REQUIREMENTS (Sakae Okubo)		S. Okubo
MPEG-TESTS (Tsuneyoshi Hidaka)		S. Okubo
MPEG-IMPLEMENTATION (Geoff Morrison)	G. Morrison

3)	The meeting selected the following input documents to the joint 
sessions from the Experts Group;

	Proposal   	: TD-7, TD-8, AVC-78 (change source to Experts Group and
			  add "QCIF" to "Resolution") 
	Information	: AVC-71,73,74,88,94,96,100,103,104
 
4)	Discussion results in ��5.2 of this report should be reflected through 
the management for and discussion in joint sessions.


7.	Others

7.1 Status report (AVC-96,105)

Updating by appointed editors is continued.  Editors are requested to send 
new texts to Chairman by September 15, 1991. 

7.2 Future meetings



    -	4th meeting 

Joint sessions with MPEG: November 18-26 in Kurihama hosted by JVC 
Sole sessions : November 27-29 in Yokosuka hosted by Japan 

    -	5th meeting joint with MPEG Video

January 7-8, 1992 in Singapore hosted by Asia Matsushita Electric
  


END


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

Annexes to AVC-106R

Annex 1  Documents for the third meeting of the Experts Group 
Annex 2  List of tape demonstrations 
Annex 3  Agreements and action points for the picture format issue 
Annex 4  Additional submission materials for the "Kurihama Tests"
Annex 5  H.26X requirements  
Annex 6  Liaison statement to SGXVIII
Annex 7  Liaison statements to CMTT/2 SRG
List of Participants of the third meeting
of the Experts Group for ATM Video Coding 

(14-23 August 1991, Santa Clara)


FRG		Mr. F. May			Daimler-Benz Research		CM
		Mr. G. Zedler	 	DBP Telecom			CM
Australia	Mr. G. Smith		AUSSAT
		Mr. P. Stevens		Siemens
Belgium	Mr. O. Poncin	 	RTT
		Mr. W. Verbiest	 	Alcatel Bell Telephone		CM
Canada     	Mr. D. Lemay		BNR					CM
USA		Mr. M.M. Anderson		Bellcore				LR-ISO
		Mr. B.G. Haskell		AT&T Bell Labs
  		Mr. M. Liou			Bellcore
      		Mr. R.P. Rao		Compression Labs
     		Mr. R.A. Schaphorst	Delta Information Systems  	CM
		Mr. A.J. Tabatabai		Bellcore				CM
		Mr. X. Yuan			PictureTel
		Mr. J. Zdepski		David Sarnoff
France 	Mr. G. Eude			France Telecom
		Mr. J. Guichard	 	France Telecom			CM
Italy  	Ms. L. Conte	 	CSELT				(CM)
Japan  	Mr. Y. Kato			Mitsubishi Electric
		Mr. T. Odaka		Toshiba
		Mr. S. Okubo	 	NTT					Chairman
		Mr. T. Soejima		Fujitsu America
		Mr. T. Tanaka	 	NTT					CM
		Mr. Y. Takishima		KDD					(CM)                     
		Mr. H. Yasuda		NTT
		Mr. T. Yukitake		Matsushita Communication
Norway	Mr. H. Sandgrind 		Norwegian Telecom			CM
Netherlands	Mr. H. Carbiere		PTT Research			LR-CMTT
		Mr. A. Koster		PTT Research       		(CM)
UK      	Mr. I. Parke		BT
		Mr. D.G. Morrison		BT					CM
Sweden	Mr. H. Brusewitz 		Swedish Telecom			CM

							  CM: Coordinating Member
							(CM): Substitute for CM
							  LR: Liaison Representative