Boulder wrote:
and when I weave the fields,the video is interlaced again!!

So when doing kvcd,ksvcd,kdvd,vcd,svcd and dvd we can't
use bob() as deinterlacer(to keep the video deinterlaced)
beacause it gives 50 fps and the standard of the above mentioned
formats is 25 fps

Or have I missed something again

--------------------
viking

We only do use (kernel)bob()) .... speratefields().selectevery(4,x,x).weave() ....in here for processing an interlaced video.
Means temporarely gettingfull height 50fps field separated streams and re-weaving the fields afterwards.

If you want to result in a finally deinterlaced 25 fps video I would do that using tomsMoComb as it works motion compensated.

Also a quick'n dirty deinterlacer is Fielddeinterlace(full=false, blend=false).

Full= false means ONLY the frames where combing is detected the deinterlacer processes, means NO unsahrp results in frames wheres no combing.

Blend=false means that no blending/ghosting will be the result, which would be my choice if fielddeinterlace() is used. But the bad side is that due the interpolation small typographics on the screen could get uneasy cause of crappy interpolation. So you have to test.

Inc. wrote:

We only do use (kernel)bob()) .... speratefields().selectevery(4,x,x).weave() ....in here for processing an interlaced video.


Ok,then everything is clear to me(finally)

Ill make some tests with FieldDeInterlace(),SmoothDeInterlace
and TomsMoComp() when deinterlacing for Mpeg 1 purposes.

Anyway thanks a lot for the help
-------------------
viking

Don't forget KernelDeint and TDeint, when considering details, they are superior

Even if I once used it ....
Drop it as its old and outdated.

So I did a test using the following script(s):

TMPGEnc settings was CQ 90,min.bitr.300,max bitr.2500,
res. 480x576,MPEG 1.

Then I burned as non-standard Vcd in Nero.
When viewing them on a 21 inch. TV,several times,the
poorest result was from FieldDeInterlace(full=false,blend=false),
a kind of "shaky" motion and not very sharp picture.It was hard to tell
any difference between the other three,maybe slightly better/sharper
picture with KernelDeint(1).

But as you can see I used default settings almost all over.
Maybe getting better results with other parameters.

BUT,the "man in the street" cant see any difference,and you must
know it to tell the difference between the results here and
the same video in Interlaced mode.Perhaps you can see
the difference better if the source was cleaner,I used
my captures for this test.

So to another question,when doing DVD,what gives best
results of this two options:

1. CQ 90,Min.Bitrate 2000,Max Bitrate 8000.

2. CBR 8000

Im here talking about interlaced videos.
And the final encoded filesize doesn't matter,only the quality.
anyone have any opinion about this?

And is CQ just as compatible as CBR with all Standalone DVD-players?
(I read somewhere that CBR is used as DVD standard,is it correct?)

---------------------
viking

I recommend the parameter type=2 for TDeint.

Also if you use TomsMoComp with the -1 parameter, you should put AssumeTFF or AssumeBFF before so that the filter gets the correct field order. I don't recommend using the vertical filter in TomsMoComp, it tends to soften too much.

If you have a truly interlaced source, you should remove full=false from FieldDeinterlace. Full=false is meant for hybrid material, in which there's both progressive and interlaced content. The other usage for full=false is being a post-processor for any IVTC filter.

Boulder wrote:
I used
Tdeint(order=1) beacause order=1 is TFF.
(Thats what I found out from the help-file)

So I guess your parameter recommandation has nothing to do with TFF/BFF?

------------
viking

That's right, type=2 means that TDeint uses kernel interpolation. The default, type=1, isn't recommended for any non-cartoon like sources. At least that's what the help file says.

@Boulder or/and Incredible
Since DGPulldown 25 > 29.97 seems to be a good method to do PAL > NTSC conversions, and it needs progressive material, could you post how would be the script to obtain progressive from PAL interlaced.
There is any change respect to the script posted in first messages of this thread?
Thanks.

http://www.geocities.com/xesdeeni200...Preview-0.html

I don't know if it will look good though.

Nice cross-entry array !

For 25i -> 25p I didn't see any problem with the method exposed, that is also the one I use (learkerneldeint)

If you can handle the extra encoding time, try tritical's TDeint. It's a very good motion adaptive deinterlacer, much less flickering but also slower than LeakKernelDeint.

http://forum.doom9.org/showthread.php?t=82264

Nice, yes, but who could explain me it?

A direct "doesnt matter" deinterlacing used on a 25i material sometimes results in a poor quality which could be better archieved by first analyzing the fieldarchitecture.

So ...

IF you got a PAL 25 interlaced source then you should go through the following steps.

1. do use after loading the source:
Telecide(order=1,guide=2,post=0)

this telecide() command in the settings above does first try to remove a dynamical phase shift on PAL material. Sometimes this fast routine lets result your PAL interlaced stream into a nice 25 fps progressive one WITHOUT blendings and thats the goal.
i.E. I did fix some SexInTheCity DVDs using this to archieve perfect 25p streams as they where just dynamical phaseshifted.

if that doesnt work, means if still combs or blendings do result, try the next one ...

2. Use one of the Deinterlacers mentioned above like LeakkernelDeint() or Tdeint(). As it seems that 25i source is a REAL interlaced one taken by a video device. BUT If the deinterlacing results in blendings then you deal with a fieldblended 25i source out of a 29.976 wich resulted from 23.976 via telecine. And this will be handled using the next option ...

2. The Restore24() or Crestore() functions (one of them).
These ones do try to inverse a process called fieldblended conversation from 23.976p -> telecine -> 29.97i -> fieldblending -> 25i.
Means if you have luck an almost clean 25p stream containing no stutters or blends will result. But these are slow, especially Restore24() and its very complex blend detection.

I would even move the Restore24 step before simple deinterlacing as such material seems to be more common than PAL video. Hopefully R24 will get faster with the new MaskTools and Didée's revision. He told me that he's at the last few steps but who knows how difficult they prove to be..

It was meant to be that I first see if a real video is the source by using a deinterlacer as second option. As if then combing still exists, this means in fact its a fieldblended source.

I tried Crestore by myself with no luck Others on Doom9 do report successes, so each one has to try by himself.

Yes, Im also waiting for the new Restore24() as it should run in LQ Mode in realtime, but even h HQ Mode is faster then the actual version.
(Theres a roumor on the germann doom9 board that Didée got rid of edgemasks as the internal blendetection in the new version will be done using a special procedure, .... so lets get surprised )

And remember that I did the same with LeakKernelDeint with no problem, except surely a time loss.
LeakKernelDeint and some others are quite "intelligent" and do not blend when not necessary.

Ok, I must have read this topic one thousand times but I still don't get it .
I just got 2 NTSC captures and I need to backup them.
Both are Satellite DVB captures and both are interlaced.
Could somebody help me with a step-by-step tour to interlaced sources ?
Please ?
Thanks in advance .

__________________
Rui

Hi Rui

Are they 24fps -->> to 29.97fps
Did you try a script with "telecide", or read here http://www.kvcd.net/forum/viewtopic....light=telecide

-kwag