Why De-Interlace?

[^AnThraX^]

I've just finished encoding a interlaced PAL-DVD (Gknot 0.28alpha, DivX 5.0.2basic).

The first time I did this DVD, (dvd2avi shows 'interlaced' in the preview,
so i used the field-deinterlace filter), the final movie looked really bad
(bad lines everywhere, like a 3d-game without anti-aliasing? dont know how to explain).
I thought, why does it look so bad?
Shouldn't I use a de-interlace filter on any interlaced source?

I encoded the same DVD again (same settings in gknot), this time
without any de-interlacing.
Result looks good, no lines at all.

So what is the real purpose of doing de-interlacing if the result is crap
and leaving it off gets you a better one?

[Awatef]

DVD2AVI shows you how the DVD is encoded, not how the movie actually is.

It is common to see progressive PAL movies encoded in interlaced mode. I know this is dumb, but as long as it looks progressive, it's not a big problem.

So, yes, don't deinterlace.
Deinterlace only if you *see* interlacing.

[^AnThraX^]

Hmm I see.

So there is no way to 'see' the real mode before I am
encoding the whole movie (or a short part of it)?

Encoding on my p3-800 takes usually more than 11hrs so
I would be very happy if I had to do this just one time :)

Anyways, thanks for the fast reply.

[jggimi]

DVD2AVI has a preview function: F5

You can create a small project file (.d2v) of a scene with movement using the [ and ] buttons. That project can then be used in AviSynth or VFAPI for frameserving to an encoder or to an AVI player for testing, as desired.

There is no reason to wait 11 hours. :rolleyes:

[JensG.]

If you move the slider in dvd2avi to places in the movie with action, you will see if it is interlaced.

[^AnThraX^]

Ah, thanks a lot :)
Problem solved.

[Awatef]

you don't even need DVD2AVI or even ripping for that, you just play the DVD in your favourite DVD player in weave mode.

BTW, you must be really unlucky to find a PAL movie being (really) interlaced! ;)

[^AnThraX^]

Hmm ok.

Yeah, guess i'm some kind of freak ;)
It was 'Black Hawk Down' RC2 PAL Rental-Edition :(

[hakko504]

I'm afraid way too many PAL DVD's are interlaced, but they should NOT be deinterlaced with a proper deinterlacer: The first action to take is Telecide(post=false,guide=2) Many DVDs (and almost ALL movies) will look correct after this simple operation.

[Awatef]

To tell the truth, I've got yet to see a PAL movie DVD that needs to be deinterlaced!
No need for Telecide or FieldDeinterlace or whatsoever for 99% of them!

[hakko504]

It seems like at least one company here in Sweden releases a lot of their DVD's as interlaced as I've encountered 5 or 6 of them in the last year, most notably E.T.

[Awatef]

You got a crappy company there, that's for sure

Anyway, you can always count on YOUR EYES to detect combing, as said, just play in weave mode, you see combing, it's interlaced, no combing, it's progressive, that's all.

[N_F]

@hakko504 - Do you have an opinion on how often telecide(post=false, guide=2) will work where Telecide(post=false) didn't?

[hakko504]

@N_F

Not really, but there is a slight speed gain from using the guide parameter, but the final result should be the same.

[crusty]

As far as I can tell, decomb automatically detects wheter or not the source is truly interlaced and simply works as a null-transform pass-through filter if it detects progressive source. So you can simply place it in your avs and it will work on whatever it finds, unless it is really crappy interlaced/ivtc/hybrid. Then you need the manual settings.

[Guest]

Originally posted by Awatef
Anyway, you can always count on YOUR EYES to detect combing, as said, just play in weave mode, you see combing, it's interlaced, no combing, it's progressive, that's all. That is not correct. You can have progressive content that displays as combed, due to a one-field phase shift (which is corrected by Telecide).

[Guest]

Originally posted by hakko504
@N_F

Not really, but there is a slight speed gain from using the guide parameter, but the final result should be the same. You can also get a speed gain from using the mm option. For example, if your matches are always c and n, use mm=2. If always p and c, use mm=1. If some p and n with c, then you have to use mm=0. mm=1 or mm=2 will be faster because one less candidate match has to be checked.

[Guest]

Originally posted by crusty
As far as I can tell, decomb automatically detects wheter or not the source is truly interlaced and simply works as a null-transform pass-through filter if it detects progressive source. So you can simply place it in your avs and it will work on whatever it finds, unless it is really crappy interlaced/ivtc/hybrid. Then you need the manual settings. Not quite accurate. Telecide() first tries to match fields, so that if there are progressive frames to be recovered (and even PAL can require it due to field phase shift), they will be recovered. But some stream conditions can result in frames that do not have a good match and those frames will emerge from field matching still combed. In that case, if you have post=true (the default) Telecide will look at each frame and try to determine if it shows enough combing to be considered combed. If it is considered combed it is deinterlaced.

It is interesting that explaining this has just given me a great idea for speed optimization! One should be able to specify a threshold for goodness of field match only below which postprocessing would be applied. This would avoid the combing check on most frames for progressive clips. The user might even optionally forego the combing check entirely for frames falling below the match goodness threshold, and just go ahead and deinterlace all such frames. That would REALLY speed things up.

What do you think of that? :)

[N_F]

Originally posted by neuron2
It is interesting that explaining this has just given me a great idea for speed optimization! One should be able to specify a threshold for goodness of field match only below which postprocessing would be applied. This would avoid the combing check on most frames for progressive clips. The user might even optionally forego the combing check entirely for frames falling below the match goodness threshold, and just go ahead and deinterlace all such frames. That would REALLY speed things up.

What do you think of that? :) Can't say I understand every bit, but as long as you do... :D

I don't know if I've said it before, but if I have it bears repeating: Thank you for all your work with varoius filters (specially decomb of course) and also for educating and helping guys like me on the forum!

[a3ternus]

Well,what a surprise to see people here talk about NON interlaced PAL DVDs...Well,I ve only seen one or two NON interlaced out of hundreds I ve rented in the past years.Unless I should NOT trust DVD2AVI when it tells me its Interlaced.Can someone cast some light over this please ?!?
A3TeRnuS

[Acaila]

Indeed, you SHOULDN'T trust DVD2AVI when it says a PAL DVD is interlaced. Instead trust your own eyes. The only PAL DVD's I have seen that are truly interlaced are Anime (and all PAL Anime are seriously messed up), all normal movies are plain progressive.

DVD2AVI has no problem with detecting NTSC, but with PAL it's almost always wrong.

[manono]

Hi-

Well,I ve only seen one or two NON interlaced out of hundreds...

I guess you'll have to go back and rent them all again so you can do it right next time. :)

[crusty]

Neuron2
Not quite accurate. Telecide() first tries to match fields, so that if there are progressive frames to be recovered (and even PAL can require it due to field phase shift), they will be recovered. But some stream conditions can result in frames that do not have a good match and those frames will emerge from field matching still combed. In that case, if you have post=true (the default) Telecide will look at each frame and try to determine if it shows enough combing to be considered combed. If it is considered combed it is deinterlaced.

Manono:
I guess you'll have to go back and rent them all again so you can do it right next time

Well from the answer from Donald I'd say that's not true, Manono.
From the answer I would conclude that on an interlaced source some frames would still come out interlaced, but on a progressive source that DVD2AVI incorrectly would detect as interlaced all frames would still be treated as progressive by decomb.
Donald, please correct me if I'm wrong or not entirely correct.
I know there are plenty of bad interlaced DVD's out there, but the issue is more about how decomb handles progressive DVD's that DVD2AVI says are interlaced.

The idea about the speedup sounds like swahili to me, but any speedup would offcourse be greatly appreciated :D

[Guest]

Originally posted by crusty
From the answer I would conclude that on an interlaced source some frames would still come out interlacedOn an interlaced source all the frames are interlaced, though you may not see manifest combing when there is no motion. For interlaced material, you would not run Telecide(), but rather a full deinterlacer, such as FieldDeinterlace().
but on a progressive source that DVD2AVI incorrectly would detect as interlaced all frames would still be treated as progressive by decomb.Telecide will recover all the progressive frames that it can, but there are several conditions that cause combed frames to sneak through. That is why the postprocessing is there.

Donald, please correct me if I'm wrong or not entirely correct.You seem to be just missing the point. I said: "You can have progressive content that displays as combed, due to a one-field phase shift (which is corrected by Telecide)." So, just because you have seen all those DVD's as combed, it doesn't mean that they are interlaced. If you meant that they appear predominantly combed even after applying Telecide(), then your experience is unique, or you are sampling from a limited pool of material, such as music videos and/or XXX material.

Just forget about what DVD2AVI tells you. Try Telecide() first. It you still see predominant combing, then you might conclude that it is true interlaced material. But better than that is to separate fields and see if there are unique pictures at each field time.

Our friend manono knows his stuff. You can trust him. :)

[crusty]

You seem to be just missing the point. I said: "You can have progressive content that displays as combed, due to a one-field phase shift (which is corrected by Telecide)." So, just because you have seen all those DVD's as combed, it doesn't mean that they are interlaced.

So your saying that progressive content can still look combed?
And I haven't seen anything combed yet, my DVD sources came out fine with the DVD2AVI progressive/interlaced choice. But I'll be more observant about that from now on.
What's a 'one-field phase shift' ?

So I should use Telecide() to check for any combing in progressive source to see if these 'one-field phase shift' occurs ? Or will it simply take out any of these phase shifts?



Our friend manono knows his stuff. You can trust him
Wasn't implying anything...:)

[Guest]

Originally posted by crusty
So your saying that progressive content can still look combed?Yes.

And I haven't seen anything combed yet, my DVD sources came out fine with the DVD2AVI progressive/interlaced choice. When it says progressive it's likely correct. But when it says interlaced, it may really be progressive. You've probably been deinterlacing them when you don't need to. They will still look fine, but they will not look as good as they could if you had recovered the progressive frames instead of deinterlacing.

What's a 'one-field phase shift'?See here:

http://forum.doom9.org/showthread.php?s=&threadid=51346

So I should use Telecide() to check for any combing in progressive source to see if these 'one-field phase shift' occurs ? Or will it simply take out any of these phase shifts? For ALL content (ignore DVD2AVI's guidance; we've told you that several times!) first just extract without any filtering. If it is not combed, do nothing. If it is combed, or has sections that are combed, run it through Telecide(post=false). If it still appears predominantly combed, then it is probably interlaced content and you should use a deinterlacer. In the case that Telecide(post=false) worked, you had phase shifts and, yes, Telecide() fixed them.
Wasn't implying anything...:) Of course not.

[crusty]

I was just wondering about a simple yet effective way to figure out if something is interlaced or not or has phase shifts (or whatever).

Goes like this:

Take two avisynth scripts and use one with telecide and one without.
Don't add any other filters.
Then encode with a constant quantizer and see which one is bigger.
If I'm correct, if there is any combing, then the one without telecide would be bigger, because combing would be harder to encode.

Am I correct in this?

[Guest]

@crusty

Perhaps. But it's neither the easiest nor the most reliable way. It's a lot easier to see combing directly that to infer its presence from the size of an encode.

[crusty]

Well I tried the quantizer methode I mentioned (used Xvid 1-pass with Q=2 for this) and it doesn't really work. The file without Telecide() is actually smaller than the file with Telecide().

And By looking at the original d2v in virtualdub in 4x enlargement frame by frame I can't see any combing.
I'm using a region 2 PAL version of 'The Thing' for this, that DVD2AVI says is progrssive. I guess then maybe it is...

I read up on the link you posted.
I'll try it again with Telecide(guide=2,post=false) to see if I get any other result.

I probably just got lucky then with all the rips I have done so far.
But I haven't done any of the sources you mentioned, like music videos or XXX stuff. Just movies.

[a3ternus]

Hmm,now this will come across as a generally stupid question but : How Am i supposed to trust my eyes on Interlaced/non interlaced PAL movies since what I use to watch them (PowerDVD 4 XP) has a built in de interlacer ?!? Maybe I should use another media player ?!? MPC handles DVDs fine without de interlacing them,right ?!?
A3TeRnuS

[Awatef]

just force weave

and please stop that whole telecide discussion, it's just non-sense! 99% of PAL movies are progressive, even if DVD2AVI says interlaced (this is the encoding method, not how the movie actually is!!).
You see interlacing in PowerDVD, use Telecide, if the result is not smooth, then use FieldDeinterlace instead, that's the whole story!! :o

TRUST YOUR EYES!!

[Guest]

Originally posted by Awatef
and please stop that whole telecide discussion, it's just non-sense! Let me get this straight. Are you accusing me of promulgating nonsense?

[Awatef]

promulgating?! what's that?! :confused:

[Mordor]

Main Entry: pro·mul·gate
Pronunciation: 'prä-m&l-"gAt; prO-'m&l-, pr&-', 'prO-(")
Function: transitive verb
Inflected Form(s): -gat·ed; -gat·ing
Etymology: Latin promulgatus, past participle of promulgare, from pro- forward + -mulgare (probably akin to mulgEre to milk, extract) -- more at EMULSION
Date: 1530
1 : to make known by open declaration : PROCLAIM
2 a : to make known or public the terms of (a proposed law) b : to put (a law) into action or force

Theres the dictionary definition of the word :)

[a3ternus]

:D