simple IVTC filter (vdub or AVIsynth) w/out decision making ?

[vhelp]

Hi all.

I hope that this is not a stupid question, so..

Is there an IVTC specific plugin or filter that does a simple
3p, 2i into 5p (progressive) output ??

What I'm on about here, is the time envolved in IVTC'ing a source
say, an known (but true) Film source ie, VHS movies. We all know
that these are pure 24 fps (or 23.976) and after we capture from
this VHS source, we get 3 Progressive frames, and 2 Interlaced frames,
all throughout the whole movie. I have found VHS source to be very
consistant in this process.
.
.
So, what I'm after here, is a direct approach IVTC filter that will
not have to make decisions on which fields/frames to hack off
and so on and so forth, but just do the IVTC process simple and quick.

If we know our source is true Film (Telecined) and not mixed as in many
DVD are, a filter that will indiscremenatly IVTC any source. That means,
if you are not sure, and you use such a filter, you would expecte
undiesireable results. But, that's ok, in the sense that we are
assuming our source is Film vs. the other option is to just use
neuron2's Telecide() / Decomb()
.
.
I mainly use neuron2's Telecide() / Decomb()
for my VHS projects, but I'd really like to use a "simple" IVTC filter
that will not make deciscions, and just parse through the source w/out
thinking :rolleyes:

If there is such a filter, I'd be happy to give it a try. At best, it
would process my source files that much faster than the above method I
use currently.

Thanks for any input on this topic all,
-vhelp

[r6d2]

Originally posted by vhelp
I hope that this is not a stupid question, so..
There are no stupid questions, only stupid answers. :)
I mainly use neuron2's Telecide() / Decomb()
for my VHS projects, but I'd really like to use a "simple" IVTC filter
that will not make deciscions, and just parse through the source w/out
thinking
I hope this answer won't turn out stupid...

Isn't just this good enough?:

Telecide(guide=1, post=false)
Decimate(cycle=5)

I know it still spends sometime on field matching, but if the source is really clean, as you say, it should get it right very often. You might try it with show=true to see if it working as expected. (I'm curious too, so please post the results.)

[vhelp]

hi r6d2,

I was assuming that this IVTC process had to be only by an external
filter, and that was the reason behind the question (and in depth
explanation in my previous post) :)

Yeah, I've ben experimeting w/ Pulldown() and SelectEvery() but
these don't seem to create the smoothness (as I thought should be
possible, IF the source is all 3p, 2i w/out any glitch or wrong
pattern, such as is the case w/ VHS movies - these are pure 3p, 2i
sources, not like DVD which are proan to mixure or bad cuts because
of the "ditial encoding and authroing" process they go through)

Any decisions as in the code snip you provided, and there is some
slow up. If you ride w/ me on my idea, then just assume that the
source is Pure and NO decision has to be made, hence the
"code" is striped out of the filter. Anyways.., I'll give you
idea a go, but I'm still debuggin SelectEvery() so far.
.
.
I've managed to get a smooth de-Interlace (aka, progressive) out of
29.970 fps source by minipulating w/ several lines of SelectEvery()
.
.
I also have a few SelectEvery() 's for 23.976 fps conversion, but
again, I'm missing something, becuase the source is jittery, weather
I field swap or jimmy around.
.
.
Still, I'll mess around some more and see where it all leads me.

Thanks for taking the time to read my posts :)

-vhelp

[Richard Berg]

http://www.avisynth.org/index.php?page=PullDown

[vhelp]

@ r6d2..

I did try your suggestion of using:
Telecide(guide=1, post=false)
Decimate(cycle=5)

but that still slowed down my encoding, even when inside vdub and
messing around w/ advancing the frames - studder movements, however.

@ Richard Berg..

I tried your suggestion at the link you provided, and although I did
read this many times (dah) I failed to realize the significance of
pulldown() however. I was using it incorrectly and w/ the wrong source.
.
.
In others words, perfect!!

I use the follwing snipplet:

doubleweave()
pulldown(0,2)

and my source was smooth. Oh, and it encoded much faster than
my previous methods :)
.
.
Again, when one knows the source well, and knows its TRUE Film, then
there's no reason to use hard-core (external) IVTC filtering, instead..,
one only has to use the above snipplet inside an AVS script and all is
well.
.
.
I will need to test this out on some more sources, but things are
looking up thus far.

Listen guys, thanks for your assistance (and reminder) :)

From the Video Workstation of,
-vhelp

[r6d2]

Originally posted by vhelp
In others words, perfect!!

I use the follwing snipplet:

doubleweave()
pulldown(0,2)

and my source was smooth. Oh, and it encoded much faster than
my previous methods :)
Gotta try that! :cool:

I've been looking for a dirty IVTC for some time, to put it in FACAR.

Thanks, Berg and VHelp.

[vhelp]

hay r6d2..

One thing though.. I haven't tested the audio part of it all :)

Yeah, I had to do a quick "rig" of my VCR to my advc-100, but only
the S-Video part of it, and not the audio. So, I have no audio in
my source, nor my testing of audio sync. I'm hoping that this will
not be an issue w/ this method.
.
.
If it proves positive, than this will be my NEW IVTC
method for Pure Film sorce (just as in Pure Interlace) hehe..

Yeah, I do hope you try it out quikly guy. I'm curious to see what
you get out of this new method (old really.., just not properly used
effectively in True Film source)
.
.
Also, remember that I said I was looking for an alternative to a
quicker method of IVTC'ing when the source is known to be
True Film. DVD (at least in my experience) has ben known to cut up
badly during scene-changes, and the Interlace fails to run through
scene-changes fluidly. So, do take that into consideration when you
test this out on what you believe to be True Film sources. I'm gonna
assume you will attempt to try this on DVD (boing) and probably fail,
casue (given the above note) these are ditigally encode and authored
badly in many cases. But, never-the-less, please do give it a try
on whatever source you are about to test on anyways :)

Ok, so far, I've tested this out on TWO test sources (VHS) and I must
say, it's smooth sailing. Now, I have to test w/ audio, but
my rig is such a mess and I don't know where/which cables go where
and to what. ..sheesh!!
.
.
Damn, this is good - no, it's great!!

Good luck thus far. Oh, yeah, Let me know how the audio fairs.. sync
or not sync.

Thanks,
-vhelp

[vhelp]

@ r6d2..

There are no stupid questions, only stupid answers.

I think Richard gave a smart answer :D

On another note..
In using the above snipplet, it may not be good for TV captures.
What I mean is, you know how badly they cut these TV shows up and
re-author them for airing. Made for TV movies are sometimes faled
or purposely jimmied. That's why Neuron2's filters are used
extensively, among other sources. But, again, I stress that as long
as you know your source for sure (through testing in vdub for instance)
for True Film, this method should not fail :D
.
.
But, this does not mean that we have to discount TV capture projects.
.
.
Take for instance, ST-E.. this show (though I've stopped watching it
since it's last season becuase they jimmy the quality to the point that
I can no long appreciate the show and so on) is done (last I checked)
(oh, and I still have a few .avi captures left scattered, for testing)
in True Film. Hmmm.. this would be a good test for TV shows. I'll
have to search these clips out. I hope that I didn't delete them. I'd
be pretty mad. Anyways..

As you can tell, I'm pretty darn excited and can't seem to stop :cool:

Ok, let me get busy. I'll be back w/ other results of my tests (and theories)

-vhelp

[vhelp]

Progress update..

Wahoo. Huston.., we have lift-off :D

Yeah.., it works on ST-E (Enterprise fans) the quick IVTC method works
perfectly, oh.., and the audio is in sync (at least w/ the lips
and on the 1 minute clip I tested it on)

Tip..
If anyone is having trouble w/ this method, tipically, you have to make
sure you are "aligned" w/ your frames. As long as you are, the video
will be smooth, and parse properly w/out blending frames (so to speak)
.
.
Also, watch out for those abrupt scene-change during commercials or
whatever gets shuved into the IVTC (w/out your knowledge) as this will
throw off your IVTC and introduce Interlace, due to a broken pattern.
So, do be careful. Source must be Pure Film (telecined) throughout
the whole source file in order for this method to work w/out issues !!

Good luck guys.

From the Video Workstation of,
-vhelp

[neuron2]

Doing a simple IVTC without any field comparisons is not a new idea. In fact it was de rigeur before the automatic tools came on the scene. The problem with it is you have to manually find the pattern phase and you have to change it when the pattern phase changes, as it often does. Handling a film with phase changes using the simple method will require a lot of time-consuming mucking about.

[vhelp]

The problem with it is you have to manually find the pattern phase and you have to change it when the pattern phase changes, as it often does.

Yes, I agree here. But, for VHS sources (at least all the ones I have
processed, I have yet to see this pattern change. My theory is that
this medium (VHS) is processed different. Perhaps, in an "analog" sense, and
not the "digital" sense, such as DVD's. These are almost always gonna have
"edits" and things, and as such, will (if not properly looked after) louzy cuts..
specially during scene changes.
.
.
In fact, to my eyes, on my ws VHS movie of "Panic Room" and during the
movie previews, on one of the movie previews (w/ the great Jody Foster)
there are cartoons mixed into the source. You know how they mess these
kinds of intermixes badly. Just like they do on TV movies that they
cut up for airing.. or those new movie previews.. perfect example was
the latest release of T3. If you recall, they cut that up pretty badly
for preview, only because it was just a preview, so it wasn't important that
they cut pricisely. Like someone's gonna capture it and encode it :eek: except me.

But remember, I'm looking for that quick method on known sources.. as
I've dramatised with VHS movies. I work w/ these lots, and I have plenty more
to do, (and even redo, ..my hobby) Perhaps many peoples no longer work w/ VHS
movies, except me. And, for me, identifying the pattern is easy.

I have yet another theory, of which I'm gonna try out first.

On another note..
I have ben cooking up an idea about coding a utility (in Delphi) to:
* open an AVI file
* read every frame
* identify all the patterns
* and produce a summary sheet or time-line of the values (ie, interlace vs. progressive)
...sort of like a snap shot graphically or something, for anylizing .AVI's - sounds
...kind of rediculous, but worth it, if not for an excercise.

Unfortunately, I don't have how to open an AVI, and read in a frame, perhaps
in a Timage, and even preview through it. That would be a milestone for me.
Anyways.. I'll have to do some searching around on google :(

Thanks for your input :)

-vhelp

[Richard Berg]

You could do that pretty easily by running Telecide's debug output through Perl/Python.

[bilu]

A primitive example that can prove helpful here:

http://forum.doom9.org/showthread.php?s=&threadid=64649

Bilu

[bilu]

@vhelp

Ever tried SmartDecimate?
http://forum.doom9.org/showthread.php?s=&threadid=60031

It's much faster than Telecide/decomb and works very well.

Bilu

[mf]

I'm surprised nobody mentioned UnComb (http://www.wideopenwest.com/~trbarry/UnComb.zip). Trbarry describes it as a pretty basic, lazy and crude filter. Just what you need? :D Pulldown is probably faster though.

[bilu]

You're right, Uncomb/Decimate is a great option :)

Although I never compared Uncomb/Decimate with SmartDecimate in terms of speed.

Bilu