[Mp4-tech] MPEG-4 compression

[Tourapis Alexandros]

Dear Steve,
You can always use the exact same encoding algorithms in an H.264 based encoder, while keeping only equivalent features as in MPEG-4 part 2 and still do a better job. For example (talking obviously for baseline) you can disable variable block sizes, multiple references, de-blocking filter (a major reason of why H.264 is more complex), intra predictions etc. Of course there are still certain things that would add up to your complexity (subpel filters, CAVLC for example) but those are definitely not increasing complexity that high, while you can do certain tricks within the encoder to reduce their complexity. This could still lead to a significant performance benefit with comparable complexity. It could even be argued, although never tested, that some of the remaining features (e.g. 8x8 within H.264) could be even removed and still maybe achieve better performance (thus possibly having even lower complexity?). 
The argument though is not on the encoder, but mainly on the decoder where, as Gary has said, a compliant decoder has to support everything.
I would argue on the fact that H.264 hardware implementations are years away. 
Alexis
-----Original Message-----
From: Steve Wright [mailto:s.wright indigovision.com]
Sent: Friday, October 31, 2023 6:30 AM
To: 'Gary Sullivan'; Steve Wright; bfelts envivio.com; Tourapis
Alexandros; Serge GEDEON
Cc: mp4-tech lists.mpegif.org; sunx
Subject: RE: [Mp4-tech] MPEG-4 compression
Hi,
Thanks, this agrees with my understanding that an H.264 codec will require
2-3 times the MIP's of an MPEG-4 codec to do a good encoding job and produce
compliant bitstreams so in many applications (eg most mobile devices) there
will not be enough processing power to provide any benefit over MPEG-4.
As H.264 hardware implementations are years away and DSP's cannot provide
enough processing power it seems that H.264 will be limited in application
to expensive high-spec PC's for now and MPEG-4 is therefore the best choice
for the widest range of applications and markets.
Steve
-----Original Message-----
From: Gary Sullivan [mailto:garysull windows.microsoft.com]
Sent: 30 October 2023 19:34
To: s.wright indigovision.com; bfelts envivio.com; Tourapis Alexandros;
Serge GEDEON
Cc: mp4-tech lists.mpegif.org; sunx
Subject: RE: [Mp4-tech] MPEG-4 compression
The basic answer to your question is that the statements about part 10
(AVC) being better than prior standards *are* made under the assumption
that a sufficient amount of processing power is available to do a good
job of using the standard.
The complexity of decoding AVC/H.264 video is higher than MPEG-4 SP
decoding.  Probably at least double the complexity.
The complexity of encoding it is up to the discretion of the encoder.
It is possible to make a very-low-complexity AVC-compliant encoder (i.e.
an encoder that does not use non-zero motion vectors) with *lower*
complexity than an equivalent-quality MPEG-4 SP encoder (because the
transform in AVC has lower complexity than the transform in MPEG-4 SP).
However, a *good quality* encoder for AVC is probably more complex than
a good quality MPEG-4 SP encoder.  Probably at least 3 times the
complexity.
A key distinction is to recognize that in order to conform to the
standard, a decoder must accept any conforming bitstream as its input,
while an encoder only needs to do what it wants to do.
Best Regards,
Gary Sullivan
+> -----Original Message-----
+> From: Steve Wright [mailto:s.wright indigovision.com]
+> Sent: Thursday, October 30, 2023 3:58 AM
+> To: bfelts envivio.com; 'Tourapis Alexandros'; 'Serge GEDEON'
+> Cc: mp4-tech lists.mpegif.org; Gary Sullivan; 'sunx'
+> Subject: RE: [Mp4-tech] MPEG-4 compression
+>
+>
+> Boris, when you say that MPEG-4 part 10 is probably the best
+> codec available
+> do you mean in the situation where unlimited MIP's are available for
+> processing? If for example you can only load a processor by
+> 200MHz is a part
+> 10 codec still better than SP?
+>
+> Thanks
+> Steve
+>
+> -----Original Message-----
+> From: Boris Felts [mailto:bfelts envivio.com]
+> Sent: 28 October 2023 21:35
+> To: 'Tourapis Alexandros'; 'Serge GEDEON'
+> Cc: mp4-tech lists.mpegif.org; 'Gary Sullivan'; 'sunx'
+> Subject: RE: [Mp4-tech] MPEG-4 compression
+>
+>
+> Thanks for your precisions Alexandros.
+>
+>   To answer Serge's questions:
+>
+> - The version of Envivio Encoding Station (2.1) released at that time
+> was using MPEG-4 part 2, more exactly it supported ASP level 5.
+>
+> - There has been few more releases of the product, still based on the
+> syntax and constraints of ASP. The current version is 2.5 and the
+> quality has improved since 2.1. Also, the quality of the
+> encoded results
+> highly depends on the set of prefilters, encoding parameters and
+> postfilters used for the considered job. This may require some
+> optimization.
+>
+> - You will find some other proprietary or standard codecs
+> which can do a
+> better job for a particular target. MPEG-4 part 2 enables a
+> fairly large
+> amount of applications and the decoder syntax has been frozen for a
+> while. Conformance insures compatible and backward
+> compatible products,
+> but may prevent from the latest technological advancement additions.
+>
+> - MPEG-4 part 10 is now part of our products, and shows significant
+> quality improvement over other legacy or proprietary codecs. It is
+> probably the best codec available. You will find (hopefully!) many
+> MPEG-4 part 10 implementations, but you need to be attentive to the
+> different profiles supported and the constraints respected by each
+> encoder to assess their quality.
+>
+> Best Regards
+> Boris Felts
+> Envivio.
+>
+> > -----Original Message-----
+> > From: mp4-tech-bounces lists.mpegif.org [mailto:mp4-tech-
+> > bounces lists.mpegif.org] On Behalf Of Tourapis Alexandros
+> > Sent: Tuesday, October 28, 2023 12:03 PM
+> > To: Serge GEDEON
+> > Cc: mp4-tech lists.mpegif.org; Gary Sullivan; sunx
+> > Subject: RE: [Mp4-tech] MPEG-4 compression
+> >
+> > According to the system's specifications seen here :
+> >
+> > http://www.visiblelight.com/mall/products/envivio/docs/ees.pdf
+> >
+> > this coding station only supports Advanced Simple Profile
+> MPEG-4, and
+> does
+> > not claim support for MPEG-4 part 10. It is claimed that both
+> realnetworks
+> > and microsoft codecs (which were developed after MPEG-4 part 2) are
+> > considerably better than this profile (but not better than
+> MPEG-4 part
+> 10).
+> >
+> > Alexis
+> >
+> >
+> > -----Original Message-----
+> > From: Gary Sullivan [mailto:garysull windows.microsoft.com]
+> > Sent: Tuesday, October 28, 2023 2:19 PM
+> > To: sunx; Serge GEDEON
+> > Cc: bfelts envivio.com; mp4-tech lists.mpegif.org
+> > Subject: RE: [Mp4-tech] MPEG-4 compression
+> >
+> >
+> > One thing to keep in mind is that standards do not specify how to
+> encode
+> > video -- only how to encode it.  There is a great deal of freedom
+> > provided in the standards that allows encoding designs of vastly
+> > differing quality.  You should never look at the quality
+> decoded from
+> > some particular implementation of an encoder using a particular
+> standard
+> > and assume that this quality is the only quality (or the
+> best quality)
+> > that is possible to obtain when using that standard.  You should
+> > definitely not expect the quality produced by all products
+> to be the
+> > same.  Not only is there freedom allowed in the design of
+> pre-processors
+> > and encoders but there is also freedom allowed in
+> post-processing and
+> > display aspects after decoding.
+> >
+> > Another aspect implicit in Sunx's response is that there
+> are a number
+> of
+> > different syntaxes that fall under the term "MPEG-4
+> video".  There are
+> > somewhere around 20 profiles in MPEG-4 part 2 and three profiles in
+> > MPEG-4 part 10.  For ordinary camera-view video, the most efficient
+> > syntax design currently in MPEG-4 is in the Main profile of MPEG-4
+> part
+> > 10 (a.k.a. AVC a.k.a. ITU-T H.264).  Serge did not specify which
+> profile
+> > was implemented in that Envivio product.
+> >
+> > And of course the RealNetworks codec that Serge referred to is not
+> even
+> > constrained by conformance to any standard.  The quality
+> it produces
+> is
+> > constrained only by the product's implementation resources, the
+> > expertise of its designers, and the deadlines of its
+> design project.
+> >
+> > Best Regards,
+> >
+> > -Gary Sullivan
+> >
+> > +> -----Original Message-----
+> > +> From: mp4-tech-bounces lists.mpegif.org
+> > +> [mailto:mp4-tech-bounces lists.mpegif.org] On Behalf Of sunx
+> > +> Sent: Tuesday, October 28, 2023 8:35 AM
+> > +> To: Serge GEDEON
+> > +> Cc: bfelts envivio.com; mp4-tech lists.mpegif.org
+> > +> Subject: Re: [Mp4-tech] MPEG-4 compression
+> > +>
+> > +>
+> > +>
+> > +>
+> > +> On Tue, 28 Oct 2003, Serge GEDEON wrote:
+> > +>
+> > +> > Dear All,
+> > +> >
+> > +> > I am currently comparing different compression solutions
+> > +> for streaming pedagogical multimedia contents over the Internet.
+> > +> >
+> > +> > But with the software I am using (Envivio coding station
+> > +> 2.1) to encode MPEG-4 content,
+> > +> > I can't reach the same quality with the same bitrate as
+> > +> with Real 9. Real 9 compression is better, the thing that I
+> > +> find a bit odd.
+> > +> >
+> > +> > Could anyone help me in this? did anyone do such a
+> > +> comparison? may it be the soft that I am using? could anyone
+> > +> advise me with another soft?
+> > +> > or may it be the parameters that I am using, I ve tryed
+> > +> aproximatly all the combinations possible!!!
+> > +> >
+> > +>
+> > +> I believe one major reason that MPEG-4 is better than MPEG-2
+> > +> is that it
+> > +> has better motion compensated coding, which is especially
+> > +> true in H.264.
+> > +> My suggestion is that maybe you need to tune those paramters
+> > +> of control
+> > +> PVOP and BVOP coding carefully. Real format is modified from
+> > +> MPEG-2 (?),
+> > +> so MPEG-4 should at least achieve the comparative performance in
+> most
+> > +> cases.
+> > +>
+> > +> >
+> > +> > Thanks in advance,
+> > +> > Serge GEDEON
+> > +> > Ph.D. Student
+> > +> > Paul Sabatier University - Toulouse, France
+> > +>
+> > +> _______________________________________________
+> > +> Mp4-tech mailing list
+> > +> Mp4-tech lists.mpegif.org
+> > +> http://lists.mpegif.org/mailman/listinfo/mp4-tech
+> > +>
+> >
+> > _______________________________________________
+> > Mp4-tech mailing list
+> > Mp4-tech lists.mpegif.org
+> > http://lists.mpegif.org/mailman/listinfo/mp4-tech
+> >
+> > _______________________________________________
+> > Mp4-tech mailing list
+> > Mp4-tech lists.mpegif.org
+> > http://lists.mpegif.org/mailman/listinfo/mp4-tech
+>
+> _______________________________________________
+> Mp4-tech mailing list
+> Mp4-tech lists.mpegif.org
+> http://lists.mpegif.org/mailman/listinfo/mp4-tech
+>
+>