&%PAGE&

-  &%page&  -
Annex 1 to Doc. AVC-206R
Annex 1 to Doc. AVC-206R

Documents for the fifth meeting of the Experts Group  
6-9 January 1992, Singapore


Normal Documents

	Note: Contributions with "*" have also been submitted to MPEG.

AVC-177R REPORT OF THE FOURTH MEETING OF THE EXPERTS GROUP FOR 
         ATM VIDEO CODING IN YOKOSUKA (CHAIRMAN)

Achievements and action points obtained at the fourth meeting are recorded 
to facilitate our discussion at this meeting.

AVC-178  (reserved for the report of SGXVIII meeting in Melbourne)

AVC-179* DEFINITION OF END TO END DELAY (CCITT EXPERTS GROUP)

All possible delays are described in a general way so that they can be 
applied to various coding schemes.  The aim is to allow a proper comparison 
between coding schemes and to give reasonable grounds for determining the 
delay budget among the elements constituting end to end delay.  

AVC-180  SUBJECTIVE TEST RESULTS OF MPEG-2, KURIHAMA MEETING 
         (CHAIRMAN OF MPEG/TEST)

This is the official report of the Kurihama subjective test for proposed 
video coding algorithms.  List of proposals, list of assessors, testing  and 
data processing methods and test results are given.   

AVC-181* EFFICIENT FLEXIBLE LAYERED CODING 
         (AUSTRALIAN "UNIVERSAL VIDEO CODEC" PROJECT)

The source of inefficiency which arise in layered coders, and methods which 
can be used to reduce inefficiency, are described.  Experimental results for 
a combined subband/pyramid scheme are presented to show that layered 
coders can be quite efficient in comparison with non-layered coders.  It is 
suggested that the general principles should be useful in the design of 
efficient layered coders. 

AVC-182  PATENT INFORMATION (DAIMLER-BENZ)

Information is given on a filed patent regarding cell loss concealment 
method.

AVC-183* CORE EXPERIMENT ON COMPATIBILITY (BT)

This document proposes a method of modifying SM3 to produce a first Test 
Model for MPEG-2 which maintains compatibility with MPEG-1 by incorporating 
a compatible prediction mode.  Items of the core experiment are also 
proposed; simulcast vs compatibility, performance of various up-conversion 
schemes, down-sampling methods to generate the SIF, improvements by use of 
motion compensation to the MPEG-1 prediction etc.



AVC-184* SIMULATION RESULTS OF PROPOSAL 30 (RTT BELGIUM)

It is informed that the submitted D1 tape suffered from field inversion 
except Mobile & Calendar at 9 Mbit/s. 

AVC-185* EXPERIMENT OF LOW DELAY MODE (GCT)

Four prediction modes (IP, IP'P, IBP and IBBP) are compared on a frame-based 
SM3 algorithm at 4 Mbit/s, concluding that the IBBP mode performs the best 
in terms of average SNR.   

AVC-186  LOW DELAY CODING MODE (FUJITSU)

Three prediction modes (IPB, IPP' and IP) are compared on a field-based MPEG-
1 algorithm at 4 and 9 Mbit/s, concluding that backward prediction 
contributes to improving average luminance SNR by 1.8 dB and 1.4 dB in 
sacrifice of coding/decoding delay.

AVC-187* CODING EFFICIENCY IN LOW DELAY MODE (KDD)

Two prediction modes are compared on a field-base SM3; IBBP with N=30, M=3 
and IP with N=15 and M=1.  It is found that IBBP performs better in luminance 
SNR by 5 dB.

AVC-188* COMPARISON AMONG COMPATIBLE CODING, NON-COMPATIBLE 
         CODING, AND SIMULCAST (FUJITSU)

The three coding modes are compared in a Hierarchical Laplacian Pyramid 
Coding scheme, providing SNR data at 4 Mbit/s and 9 Mbit/s.  The layered 
coding looks effective at 4 Mbit/s, but it performs similarly to the 
simulcast coding at 9 Mbit/s. 

AVC-189* COMPARISON BETWEEN PYRAMID (EMBEDDED) AND SIMULCAST 
         (TOSHIBA)

The two coding modes are compared in an RM8-based coding scheme where the 
base layer has either of CIF or QCIF resolution.  It is found that the 
simulcast mode performs better than the pyramidal mode supposedly due to 
the fact that motion compensation becomes less effective for the 
differential signal in the pyramidal scheme.

AVC-190* CONSIDERATIONS ON CELL LOSS RECOVERY TECHNIQUES (JAPAN)

The following five techniques for the cell loss recovery are detailed; leaky 
prediction, demand refresh, error concealment, video multiplex structure, 
and packetization.  Items of standardization are pointed out.  

AVC-191* RATE CONTROL METHOD BASED ON SUB-BAND CODING FOR VBR 
         TRANSMISSION (OKI ELECTRIC)

Experimental results are reported on a control method for maximum coding 
rate, measured with 1 frame time jumping window, in a 7 sub-band coding 
scheme.  The control method is to start with the previous frame quantizer 
step sizes and adjust them in the order of UU, UL, LU, LLUU, LLUL and LLLU 
band.  

AVC-192* COMPARISON OF PREDICTION/CODING METHODS IN AN RM8-BASED 
         SCHEME (NTT)

Three prediction/coding methods (frame, field and adaptive) are compared in 
an RM8-based coding scheme in terms of rate vs distortion characteristics.  
It is concluded that frame coding with simple field merging is a reasonable 
starting point and that if field mode is the choice two preceding fields 
should be used for motion compensated prediction. 

AVC-193* OBSERVATION OF SEQUENCES WITH CYCLICALLY DISTORTED 
         PICTURES (NTT)

Flower Garden sequence is processed by inserting intra coded pictures with 
constant step size in every other frame.  Just noticeable degradation is 
given for the three parts; flowers, roof tile, and twigs.  A possibility of 
exploiting cyclic coarse quantization is suggested but not firmly concluded.  

AVC-194* FIELD ADJUSTED MC FOR FRAME-BASE CODING (MATSUSHITA)

A new motion compensation method is proposed for the frame-based coding of 
interlaced pictures, hence for TM0.  A motion vector is sent per MB, from 
which a field-time difference compensation vector is derived.  Odd (even) 
field pels are MC-predicted by the picture at the previous odd (even) field 
position that is composed of the original previous odd (even) field pels and 
even (odd) field pels shifted by the field-time difference compensation 
vector.  

AVC-195* CODING EFFICIENCY COMPARISON BETWEEN FRAME FORMAT AND 
         FIELD FORMAT (HITACHI)

Coding efficiency is compared between three frame-based coding (frame DCT, 
field DCT, field/frame adaptive DCT) and one field-based coding in an SM3-
based coding scheme.  Used test sequences are Mobile & Calendar (panning) 
and Table Tennis (zooming out).  It is found that there is small difference 
among the three frame-based coding and that frame-based coding performs 
better for Mobile & Calendar but field-based coding performs better for 
Table Tennis.  

AVC-196* GUIDING PRINCIPLES FOR THE SINGAPORE MEETING (CCITT EG)

Some guidelines are proposed for defining TM0; starting from a very simple 
scheme for later inclusion of verified elements, room for optimizing low 
delay mode, core experiment for backward/forward compatibility with MPEG-1 
and H.261, room for forward prediction only, scalability and low bit rate 
operation.  

AVC-197  SIMULATION OF RANDOM CELL LOSS (UK)

A method of simulating cell loss is given, where bitstream specification, cell 
loss occurrence model, mean cell loss rate and mean burst of consecutive 
cells lost, programs for cell loss and random number generation are 
described.  Parameter values are proposed for common experiments.

AVC-198  CORE EXPERIMENT FOR TESTING B-FRAMES/NO B-FRAMES (NTA)

It is proposed to compare "B-frames as in MPEG-1", "low resolution predicted 
frames in prediction loop", "low resolution predicted frames out of 
prediction loop", and "prediction based on two previous frames".  
Experimental conditions are also proposed.

AVC-199* A STUDY ON THE EFFECT OF B FRAME (NEC)

Comparison of coding efficiency is made between IBP and IP structures, 
concluding that the former gives better SNR results.

AVC-200* A STUDY ON SUBSAMPLING OF CHROMINANCE SIGNALS (NEC)

Four chrominance subsample methods to obtain 4:2:0 signals are compared in 
terms of SNR and subjective impression; line subsampling and 
repeating/filtered interpolation, filed drop and repeating/filtered 
interpolation.  It is concluded that a method performing no worse than the 
line subsampling and filtered interpolation is required.

AVC-201* SUMMARY OF FUNCTIONALITY PROPOSALS 
         (CHAIRMAN OF MPEG/REQ)

Description of functionalities (fast replay, random access, compatibility, low 
delay mode, scalability, cell loss resilience) is listed which appear in the 30 
proposal documents.

AVC-202  SELECTION OF TRAFFIC DESCRIPTORS AND THE IMPACT OF 
         BUFFERING (AT&T)

Distinction between TD (Traffic Descriptor) and UPC (Usage Parameter 
Control) is first given.  An algorithm to select parameters for TD's (sliding 
window and leaky bucket) is next provided such that the UPC will find no 
violations, demonstrating that an additional peak rate constraint is 
advantageous and that the leaky bucket is a better choice.  Furthermore, it 
is shown that buffering in the video system can reduce the necessary 
parameters.  

AVC-203  CONVERSION FILTER FOR PRODUCING SCIF (NTA)

A filter is defined to convert from either 525/60 or 625/50 signal to SCIF.  
Tape demonstration is accompanied.

AVC-204* EVALUATION OF SM3 FOR MPEG2 (PTT RESEARCH)

It is reminded that the following requirements should be considered for 
appropriate actions before starting with a modified SM3 as TM0; end to end 
delay, error protection, adaptation to various storage and transport 
methods, and compatibility. 

AVC-205* CELL LOSS EXPERIMENT SPECIFICATIONS (EXPERTS GROUP)

This is an update of AVC-197 with some modification in the bitstream 
specification and parameter values.


Temporary Documents

TD-1	Agenda for the fifth meeting in Singapore (Chairman) 
TD-2	Available documents (Chairman)


END