&%PAGE&

-  &%page&  -
Annex 3 to Doc. AVC-256R
Annex 3 to Doc. AVC-256R


Summary of discussion on the picture format issue

Small meeting on Video Format 19 March 1992

Present:	
G Bjoentegaard
L Conte
B Haskell
F May
G Morrison (Chair)
S Okubo
D Schinkel
G Smith
Y Takishima
C Verreth

At the Santa Clara meeting in August 1991 the Experts Group had determined 
to make a decision on video formats at the earliest opportunity in 1992.

One proposal was the Super Common Intermediate Format (SCIF), being an 
upscaled version with twice the number of pixels in horizontal, vertical and 
temporal axes of CIF from Recommendation H.261.  The SCIF concept would 
provide the guaranteed compatibility between all equipments including 
connections between 525 and 625 line regions in the same way as CIF in H.261.  
The other benefits of more commonality in equipments and ease of multipoint 
working would also apply.

Another proposal offered much more flexibility of the internal format used 
for coding.  A modest number of classes would be defined, characterised by 
maximum numbers of pels in the three axes.  Decoders belonging to a class 
would be able to decode any image smaller than the relevant maxima.  
(Spatial dimensions could be restricted to integer multiples of 16 to fit the 
macroblock configuration.)  In this proposal the issue of possibly different 
capture and display formats (pels per line, lines per picture, pictures per 
second, pel or picture aspect ratio) would be handled entirely by the 
receiving terminal.  Coding would take place in the originating scanning 
format thus avoiding any quality degradations from unnecessary standards
conversion in connections where the display format was the same as the 
originating format.  Conversion might or might not be employed in receiving 
terminals after decoding in connections where display and originating 
formats were not the same. 

These two proposals can be regarded as the extreme ends of a range with 
many others in between.  These would comprise a number of defined formats 
available for use in the coding kernel, probably comprising the natural 
formats of the major applications requiring the highest quality.  Other 
source formats could be converted before coding to the nearest suitable 
one of the set of defined formats. Receivers would display the decoded 
version directly or after conversion to any other required format. 

For the March 1992 meeting much work had been done on the proposed SCIF, 
including simulations of various conversions and coding performances.  
Facilities were available at the meeting for viewing all the relevant 
formats, some for the first time such as SCIF directly.  After these informal 
assessments the Experts Group agreed that:

1.	Line number conversion can be achieved with small or invisible quality 
loss.

2.	Conversion from interlace to progressive format is not totally 
satisfactory from a quality viewpoint though several methods have been 
tried.

3.	Picture rate conversions are almost always accompanied by visible 
defects.

Thus it was concluded that the use of SCIF in circumstances requiring 
conversions other than of line number cannot, at least with the methods 
tried, yield the picture quality expected of H.26X at bit rates in excess of 
about 5 Mbit/s.

One demonstration where the progressive format was used throughout the 
capture, coding and display processes was acknowledged to provide very 
pleasing pictures and showed the potential benefits of progressive over 
interlaced formats.

A contribution at the March 1992 meeting pointed out the approach adopted 
so far by ISO/MPEG would result in their video coding algorithm being able 
to cover all the formats discussed in the CCITT Experts Group.  The Experts 
Group anticipates that H.26X will be the same or fundamentally the same as 
the MPEG one.  Therefore, a decision by the Experts Group would not hasten 
the algorithm development and can be deferred.

The Experts Group agreed:

1.	to defer a final decision on picture formats to be specified in H.26X.

2.	to continue to use both 525/60 and 625/50 versions of CCIR-601, and 
where possible the related Extended Definition (EDTV) and High Definition 
(HDTV) formats, in the development of the H.26X coding algorithm.

3.	to continue investigation of format conversion methods.  In some 
circumstances these will be unavoidable and availability of satisfactory 
methods is highly desirable even if not subject to standardisation.  
Guidelines are listed at the end of this annex.  (More advanced methods 
exist, such as motion compensated techniques, though they may be uneconomic 
for widespread use in terminals.)

4.	to study application scenarios to identify issues which really need 
solutions and the performance targets which should be met, and try to have 
a common understanding by the London meeting (November 1992).

5.	to be aware that square pixels are utilized in some potential 
applications.  (Currently displays with square pixels are available, but the 
corresponding digital acquisition equipment is not.)

6.	to be aware that formats other than those from the television 
industry may be applicable for some applications.  An example is computer 
displays.

7.	that the specification of complete systems giving interworking is 
required by CCITT.  Though it is not yet clear whether the format issue will 
be dealt with entirely by the video coding Recommendation H.26X or entirely 
by the terminal Recommendation H.32X or by both in combination, the 
expertise of the Experts Group is needed by SGXV.

8.	to recognise the potential of progressive formats and endeavour to 
support them in H.26X/H.32X for eventual use sooner or later.


*   *   *

Requirement guidelines for format studies

1.	Degradation from standards conversion must be consistent with 
intended use.

2.	Any loss of coding efficiency caused by standards conversion must be 
acceptably small.

3.	Delay introduced by standards conversion must be acceptably small 
for the intended use.

4.	Equipment complexity overhead must be acceptable.  More study is 
required to determine the true impact of formats with higher numbers of 
pels. 

5.	In circumstances when standards conversion is required there are the 
two approaches of going directly from one to the other or of going via a 
third (intermediate) format (such as SCIF).  The two approaches should be 
compared. 


END