Decomb()Killing interlacing Titles

[pila13]

Hello, problem with which I can not understand -
I use Decomb to deinterlacing animations, all fine..but cannot "kill"
interlacing on titles -

http://z1.zod.fr/z/1-ig7.png
http://z1.zod.fr/z/2-jg7.png
http://z1.zod.fr/z/3-kg7.png

Use this script -


edeintted = AssumeTFF().SeparateFields().SelectEven().EEDI2(field=1)
tdeintted = TDeint(edeint=edeintted,order=1)
Telecide(order=1,guide=0,gthresh=8,chroma=false,show=false,blend=false,post=2)

Try use diffirent post/guide/gthresh e.t.c, Don't help.

If some one have ideas please let me know

:thanks:

[Guest]

I don't see any interlacing on the titles in the screenshots you posted.

They are presumably *deinterlaced* and you are complaining about the effects of deinterlacing. Is that correct?

You should post an unprocessed source sample so that we can duplicate your issue.

Your script makes no sense to me: deinterlacing and then field matching!?

[pila13]

unprocessed (http://www.mediafire.com/?junjonxt1dz)
use decomb (http://www.mediafire.com/?ymmjm0ynjm3)

Your script makes no sense to me: deinterlacing and then field matching!?
What do you mean ?
Sorry but m english not wery good...

[Guest]

Why are you giving me an AVI as an "unprocessed" source file? Where did you get it?

I want a much larger sample of the VOB directly from the disk. You can use DGSplit to cut it. Alternatively, set a range in DGIndex and then "save project and de3mux video"; upload the resulting M2V.

[pila13]

I would like to make a small file on the contrary ....
Clean sample (http://file.qip.ru/account/file/133687508/35116009/3131demuxed.html)

[Guest]

The sample you posted is not interlaced! It is progressive with a one-field phase shift, which you can prove with this script:

dgmultisource("3131.demuxed.dgi") # or mpeg2source("3131.demuxed.d2v")
separatefields().trim(1,0)
assumetff().weave()

There is some residual flicker but that is present in the source (see below for a possible reason for it).

I repeat my point about your script making no sense but since you have only given me the credits section I cannot make any suggestion for you about handling the main movie.

That is unusual material. It almost looks like the credits were bobbed from a lower rate. What is the title and region of the DVD?

[Stephen R. Savage]

EDIT: Previously there was incorrect information here.

[Guest]

However, I'm going to have to disagree with your assertion that the clip is phase-shifted progressive. A full-rate bob shows unique motion on each frame. The motion is large between frames and very small between the fields (after adjusting the phase). That interfield motion looks like the remnant of a previous bobbing operation. It certainly is not classical interlacing, because the motion would be constant between any pair of fields, and that is not the case.

To see this, separate the fields and step through them.

[Stephen R. Savage]

If that were the case, then why is the motion uniform in both direction and speed when I step through the bobbed frames? If you look at a small vertical scroll like this through separated fields, wouldn't you be let astray by the 1px shift between odd and even fields?

[Guest]

If that were the case, then why is the motion uniform in both direction and speed when I step through the bobbed frames? If you look at a small vertical scroll like this through separated fields, wouldn't you be let astray by the 1px shift between odd and even fields?

It happens because you are not correcting the phase shift.

Step through the fields as I said. There is no arguing with it.

P.S. Since you deleted the post I responded to, leaving my reply looking orphaned, I restored your post and quoted it. I think you have a possible ego problem because you regularly delete posts when you are shown to be misinformed or wrong about something. A mature person would say something like "I see, thank you for the clarification." There is also a possible element of sociopathy because you don't care if your deletions leave your respondents' posts thereby orphaned and/or unintelligible.

[Overdrive80]

Hello, problem with which I can not understand -
I use Decomb to deinterlacing animations, all fine..but cannot "kill"


You can use option VFR of animeivtc. Regards

[Stephen R. Savage]

It happens because you are not correcting the phase shift.

Step through the fields as I said. There is no arguing with it.

P.S. Since you deleted the post I responded to, leaving my reply looking orphaned, I restored your post and quoted it. I think you have a possible ego problem because you regularly delete posts when you are shown to be misinformed or wrong about something. A mature person would say something like "I see, thank you for the clarification." There is also a possible element of sociopathy because you don't care if your deletions leave your respondents' posts thereby orphaned and/or unintelligible.

Actually, I deleted my post because I felt the discussion was not progressing anywhere useful. Since you force me back into it, I will point out that I do not believe I am wrong. I believe the reason switching the fields works is that the scrolling motion is on the same order of magnitude as the odd/even shift, and by reversing the field order, you are performing a crude version of motion compensation. This can be seen in 1) the fact that the image quality is highly degraded afterwards, and 2) the much sharper result one can obtain by using TempGaussMC, which produces uniform vertical motion at 60fps.

For example, consider an object moving vertically at a rate of 2 px/frame. On the first frame, it would occupy lines 1-2, then 3-4, etc. After interlacing, it would occupy field 1, then field 4, then field 5, then field 8, etc. If you reverse the field order or delete a field, the apparent 3 line shift would become a 1 line shift, producing the result you describe.


Lines Field Reversed Weaved
1-2 1 2 2-3
3-4 4 3 2-3
5-6 5 6 6-7
7-8 8 7 6-7

[Guest]

Actually, I deleted my post since I felt the discussion was not progressing anywhere useful. Sure, that's why you edited it to read like this (though likely you'll change it again after seeing this):

EDIT: Previously there was incorrect information here.

Your new explanation is also incorrect. I'll leave it to the readers to judge the matter now, because you appear unable to comprehend the simple demonstration of your error that I presented (step through the fields and observe that there is not constant interfield motion, and the big motions are certainly not "on the same order of magnitude as the odd/even shift" as you erroneously claim).

[Stephen R. Savage]

Actually, stepping through the fields, I can only confirm exactly what I wrote. Fields 1-2 appear to be in roughly the same position when looking at the separated fields, which means that they are actually 1 px apart spatially. Vertical motion between fields 0-1 is on the order of 3-4 px, as I described in my post. I can not see how this does anything but confirm my hypothesis. As a senior member of Doom9, I would have thought you to understand that what appears to be uneven motion in fields can very well be smooth motion in the original frames, as the same line in odd and even fields do not correspond to the same line in a frame.

However, I know that you will resort to making yet another ad hominem attack. And, since you are a moderator, I will be forced to concede your point, even if it is incorrect. In fact, I had previously attempted to hand the point to you, but you refused my offer. I have no further desire to try to convince you, as the task is fruitless.

You say that you will allow the readers to decide, in which case I will present my evidence in the form of easily understood images:

Removing the first field and switching field order:
http://img694.imageshack.us/img694/7863/fieldswapping.png

Deinterlacing:
http://img693.imageshack.us/img693/1705/deinterlacing.png

If anyone doesn't believe me, they are welcome to compare scripts and try for themselves:

Field switching:

MPEG2Source("3131.demuxed.d2v")
Trim(1, 0)
AssumeTFF()
Weave()


Deinterlacing:

MPEG2Source("3131.demuxed.d2v")
TempGaussMC_beta2()
SelectOdd()


P.S. I do not appreciate at all your insinuation that I edit my posts to mislead others. As a moderator, you can very well see that I only edit posts to fix errors and add additional information, except in this case earlier where I had attempted to let you save face.

P.S.2. Rest assured that this will never happen again, as I have learned the consequences of attempting to argue with a moderator. Have fun in happy-land.

[pila13]

neuron2 and Stephen R. Savage
Thx you!