[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[OpenDivX] OpenDivX Encoder -> OpenDivX Daily Forums Digest

Topic:		scene-change / fade detection
Author:		gruel
Posted:		 2001-03-29 04:39
I noticed that the scene-change mechanism does not work well on
fadeovers (same with Divx3.11 btw). They are often not recognized which
leads to much data and still a very blocky picture (trying to build a
B-frame without a matching reference, I guess).

I read in a paper from IBM about fade-in/out and fadeover detection and
setting motion vectors to zero there, thus having more bits for the
frame itself. Is anybody working on these things?

Btw. could someone of DARC make a list of what is worked on by whom?
E.g. "bitrates controll" by JoeCool; "motion estimation" by JohnSmith,
"general speedup" by MrClean...
Or are these things not distributed among members? 

Topic:		scene-change / fade detection
Author:		eagle
Posted:		 2001-03-29 05:10
We are aware of the problem with fades, most simple SAD-based ME schemes
are confused by fades and crosfades.  Even my MPEG2-based digital cable
TV looks sh!t when there is a fade!  We have a number of ideas for
fixing this and our next release should be improved in this respect.

If you have any original solutions we would be glad to hear them.
Please do experiment.

Regarding who in DARC is doing what - it varies all the time.  If you
have a specific query mail me (or one of the others) and I'll forward
it.  All of us are very very busy!

Topic:		What can be expected 'bout realtime encoding?
Author:		SVB
Posted:		 2001-03-29 07:38
You would be better to forget it, because you need much stronger machine
to capture at realtime. If you capture with these settings at realtime,
you will get really ugly quality or really large files, nothing matter
which codec you choose.

Topic:		Variables Farbprofil
Author:		SVB
Posted:		 2001-03-29 07:53
You have completely misunderstood the problem.
The color information uses only about 1/2 of space of luminance
information, even when uncompressed. So you would theoreticaly get to
66% of original file size, if you ommit colors. But that is true only in
uncompressed case.
When you compress b&w video with mpeg4, codec "sees" that there are no
colors and thus uses much less space for storing them.
So if you would write codec for b&w video coding, the space gain is just
about few %.

And the other thing: you cannot compress paletted videos (2, 16, 256
colors) with the same algorithm as normal video. You would have to use
completely different principles.
Compression of such videos is commonly done by converting them to
standard 24bit color depth and there are good reasons to do so.[ This
message was edited by: SVB on 2001-03-29 07:53 ]

Topic:		Variables Farbprofil
Author:		Duzmi
Posted:		 2001-03-29 09:26
Ich weiß, daß es für heutige Dekodierungsverfahren
so gut wie unmöglich ist, ein Video mit mehreren
Farbprofilen gleichzeitig zu dekodieren.
Doch darin liegt ja eben der springende Punkt:
Gelingt es dem DivX Codec (OpenDivX 4.0 alpha49),
das Video in mit mehereren Farbprofilen zu
dekodieren, wäre das ein neuer Schritt...

Und man kann mir nicht sagen, daß es nicht möglich
ist, einen Code für dieses spezielle Verfahren zu schreiben.
Heutzutage läßt sich doch alles machen!

Es würde ohnehin leistungsfähigere Computer erfordern,
doch die Hardwareanforderungen für heutige
Aufgaben am PC steigen von Tag zu Tag mehr.
Heutzutage ist ein Pentium III 550 MHz absolutes Minimum
für Videobearbeitung (sprich MPEG-2 Kodierung).[addsig]

OpenDivX mailing list
[email protected]

Reply To Poster

Local References / HOW-TO / FAQs