IVTC - What is the ivtc 'number pattern' for TMPGenc ?

[vhelp]

Hi guys

I just got my Analog Cable, and I really love it.

I'm a very happy camper, hehe

Question:
I got this source (cartoon) and it in something like 4:1 pattern, (instead
of the usual 3:2 pattern) and in TMPGenc, when the pattern is 3:2, I will
use TMPGenc's IVTC filter, and enter in the numberical value: 10100

This will give me a perfect IVTC (or 23.976 fps) conversion.
But *how* do you do this for a 4:1 pattern ? I'm tried all kinds of
numbers, but I just can't figure it out.

The pattern goes like [n] this..

PPPP:i PPPP:i PPPP:i PPPP:i PPPP:i


The usualy 3:2 pattern [10100] is..
PPP:ii PPP:ii PPP:ii PPP:ii PPP:ii

Thank you for any assistance on this.
-vhelp 3797

[jagabo]

You have 2:3:3:2 pulldown. I don't have any video like that here but I think the IVTC pattern for TMPGEnc 1010101000, starting on the first P frame.

[vhelp]

@ jagabo, evening..

Thanks for your help earlier. I appreciate it. Wish I could really
test it out on a fluid and clean source, but the source that I did
try your 1010101000 ivtc code did not work, but only because
the source film (cartoon) was eratic. It did work for within one
scene. But, just not throughout the whole source

(And how you were able to figure out that the source was 2:3:3:2 pattern
is beyond me - great work)


May I trouble you again ...

But, now I have another source who's IVTC I just don't recignise.
And, because I use TMPGenc extensively, (not avisynth) for its RGB
requirements, I need additional help in determining the source type
IVTC pattern. This one is pretty crazy, IMO.



** Pattern: Pi IIII Pi IIII Pi IIII Pi IIII

where the small " i " seems to be a semi-blend, while the large " I "
are regular Interlace. I could be wrong. But this is the best I can
do at explaining it.


If you would like to test this one out I can provide a 48 frame source..
DV source from my ADVC-100 and my currently new, Cable source. The source
program is an English tv program from channel Thirteen.

(it is really tricky, and I just can't get TMPGenc to ivtc it (or I can't,
that is) --

** Also, can you demonstrate how to determin this inside TMPGenc's ivtc
** editor window ??
**
** Or, do you know of a How-To on using TMPGenc's ivtc editor to determine
** ALL types of ivtc patterns ?? (feel free to discuss via PM, if you wish)

Thanks again for your help earlier (sorry I didn't get back to you then)
..much is greatly appreciated

EDIT: - added a .RAR file of the frames.
** DOWNLOAD: masterpiece.theatre.rar - size is 5.1MB

-vhelp 3801

[jagabo]

the source that I did
try your 1010101000 ivtc code did not work, but only because
the source film (cartoon) was eratic. It did work for within one
scene. But, just not throughout the whole source :(

That's often a problem. Usually because a video is telecined first, then edited. Or sometimes because of dropped frames during capture.

But, now I have another source who's IVTC I just don't recignise.
And, because I use TMPGenc extensively, (not avisynth) for its RGB
requirements, I need additional help in determining the source type
IVTC pattern. This one is pretty crazy, IMO.



** Pattern: Pi IIII Pi IIII Pi IIII Pi IIII

where the small " i " seems to be a semi-blend, while the large " I "
are regular Interlace. I could be wrong. But this is the best I can
do at explaining it.


If you would like to test this one out I can provide a 48 frame source..
DV source from my ADVC-100 and my currently new, Cable source. The source
program is an English tv program from channel Thirteen.

(it is really tricky, and I just can't get TMPGenc to ivtc it (or I can't,
that is) -- :(

The sample video you provided was an interlaced 25 fps PAL video converted to interlaced 30 fps NTSC. You won't be able to IVTC it since the original source wasn't progressive.

** Also, can you demonstrate how to determin this inside TMPGenc's ivtc
** editor window ??
**

It's a little difficult to describe. Let make sure you understand what TMPGEnc is showing you in the IVTC dialog. I'll use top-field-first in this example:

Along the top if the dialog are thumbnail images. These show each possible combination of fields into frames. Thumbnail 0a is showing the combination of the top field of video frame 0 and the bottom field of video frame 0. Thumbnail 0b is showing the bottom field of video frame 0 and the top field of video frame 1. Thumbnail 1a is showing the top field of video frame 1 with the bottom frame of video frame 1. THunbmail 1b is showing the bottom field of video frame 1 with the top field of video frame 2.

For example, consider a series of video fields as broacast. Each digit represents a field:

0123456789

Your SD TV simply displays each field as it's recieved. Let's say all the even numbered fields are top fields and the odd numbered fields are bottom fields. When captured by a computer the fields are combined into TFF frames. Each pair of digits represents two consecutive field woven together into a frame:

01 23 45 67 89

TMPGEnc's thumbnails are showing the following combinations of fields (let's call them "field pairings"):

01 12 23 34 45 56 67 78 89 90

The 0's and 1's in TMPGEnc's IVTC pattern are simply bit masks indicating which of those field pairings to include in the output. So a pattern like 10100 (the usual pattern to IVTC 3:2 pulldown) says to include 01, exclude 12, include 23, exclude 34 and exclude 45. Then the pattern repeats.

So figuring out the appropriate pattern is a matter of assigning all the progressive pairs a 1, and all the interlaced pairs a 0.

This is complicated by the fact that videos often fade in from black or have no motion at the start. So you have to go somewhere in the middle of the video to find obvious motion, figure out the pattern, then figure out how to apply the pattern at the start. Ie, that basic pattern to IVTC a 3:2 pulldown video is 10100, but it might start with any one of the five positions:

10100
01001
10010
00101
01010

It is further complicated by the fact that many different pulldown patterns can be used. 3:2, 2:3:3:2 are common for film to NTSC. 3:2:3:2:2 is often used for progressive PAL to NTSC. So you have to figure out the length of the pattern too.

[jagabo]

vhelp, I was looking again at the Masterpiece Theatre clip. It DOES appear to be from an progressive 25 fps PAL source. The partially blended fields had me fooled for a bit.

It's 3:2:3:2:2 pulldown (PAL to NTSC). The IVTC pattern for TMPGEnc starting at field pairing 0a is 101010010010. This will give you a 25 fps result so you'll have to specify 25 fps for the output in TMPGEnc. You're progressive result will still have some ghosting from the partially blended fields.

Your DV AVI is bottom-field-first of course. Not that it particually matters here.

I don't know where you're going with this but I'd be inclined to just leave it as interlaced video if you're burning an NTSC DVD.

[manono]

Hi-

Geez, people still use TMPGEnc for IVTC? Your earlier one could have easily and much more quickly been done using the Decomb or any other IVTC filter in AviSynth. This one, as jagabo says, is from a progressive PAL source and blended to 29.97fps interlaced. Using TMPGEnc to IVTC it will leave most of the ghosts/blends/double images and also make it play jerky. Using the RePAL filter in AviSynth will remove the ghosts (almost all anyway), and return it to the original framerate (well, 24.975fps anyway). Here's the script:

LoadPlugin("C:\Path\To\LeakKernelDeint.dll")
LoadPlugin("C:\Path\To\RePAL.dll")
AVISource("C:\Path\To\Masterpiece.avi")
ConvertToYV12()#or YUY2
LeakKernelBob(Order=0)#change to 1 if it's TFF for the entire clip
RePAL()

Here's a link to the RePAL'd sample (1.16 MB) done at quant 3 using XviD:

http://rapidshare.de/files/12205620/RePAL.avi.html

Let me (or anyone else that knows about this) know if you want to pursue the AviSynth solution and have any problems.

[jagabo]

RePAL worked pretty well on the short sample (I reproduced your result here). It was much cleaner (ghost free) than TMPGEnc's IVTC.

[vhelp]

Hi all

I'm sorry I haven't responded sooner. I've been pretty busy with a lot
of things on my plate.

I had problems with my new Analog Cable. Apparently, I learned some
conspiracy (Digital vs. Analog) that I had uncovered, and I had to fight
to get my Analog. I don't want "digital" as its just too blocky. And
with the 2006 Olympics just a few short days away, (Torino, here I come)
You DEFFINATELY DON'T want digital cable/satellite, unless you like a bunch
of square pixels. And, to add more cakka to the nonsense, it will be in
pure interlace - worse for digital - lots more pixelations. sheesh.

So, I've ben pretty busy getting ready. Anyways.

(to be honest, I like to read other peoples problems/posts and respond
back if I can. I know.. pretty crazy and selfish, but I don't mean anything
by it - really. By then, I'm exhausted)


I wish I could redo (capture) that Masterpiece video again, cleaner than
it was, but I don't have a tv guide, and don't know when it comes on again.
I think it's every Sunday.. but I don't know about if it repeats, that is.
But, I guess its not important. I will have other Sundays to capture this
(what seems to be, PAL->NTSC video project - something that I thougth I
would *never* get into here, mixed sources.. oh well)

@ manono

I will try your script. I don't have the necessary plugins. So I will
have to search those out, then try them. Maybe I can refine the script
to do a better job - not to say that your's was wrong. Just a natural
reflex of mine.. to enhance, improve or make better Anyways.

I might try my hand at the complete source, just to see how it goes, through
the whole PAL->NTSC process. I'm curious

And to be honest, had never thought to consider weather this source was
a PAL->NTSC conversion project. It was the furthest thing on my mind.
I just assumed it was a weird NSTC ivtc pattern that had to be cracked
open - but with TMPGenc. I was wrong.

Thanks guys for your input. I will consider both methods.. though I am
really interested in the TMPGenc ivtc editor method, because that is what
I am most interested in, hence the purpose of this thread/topic.

@ jagabo

I will study that mini tutorial you posted - thanks.

-vhelp 3805

[manono]

Hi vhelp-

You can find both LeakKernelDeint and RePAL here:

http://www.avisynth.org/warpenterprises/

As for refining the script; I used the fastest good Smart Bobber (LeakKernelBob). For slightly better quality, you might try the bobber within TDeint. It can be found at the same place, although I think they have an older version. For the latest versions of tritical's filters, check the bottom of this post:

http://forum.doom9.org/showthread.php?p=544121#post544121

It's somewhat slower. And for the best quality bobber, but excruciatingly slow, then you'll want to use EEDI2 in bob mode, also at that second link. For unblending, there are other filters and functions, such as Restore24. Personally, I've never found anything to beat RePAL, and I see this kind of thing all the time, and have used it dozens and dozens of times. I'm using it right now to encode 7 extras and a main feature from a PAL2NTSC DVD.

If you eventually plan on encoding to DVD with TMPGEnc, then you'll want to add ConvertToRGB24() to the bottom of the script. Since TMPGEnc is slow enough to begin with, you might want to encode to a lossless AVI codec first, and then feed that to TMPGEnc. I do that with multipass CCE encodes. It may or may not be worth it for a 2-pass TMPGEnc encode.

And then, when all done, you can keep the audio untouuched if you like by using the custom setting of DGPulldown for 24.975->29.97 (if NTSC). Or slow the framerate using AssumeFPS(23.976) and run regular pulldown. That will, of course, also mean slowing the audio.

In the furure, to help find if you have a PAL2NTSC blended source, put on the smart bobber by itself. If you see the blended fields, then that's a pretty good indication that it's PAL2NTSC. There can be other reasons that the fields get blended, but that's by far the most common. In those cases, RePAL is your friend, as no regular IVTC will do it correctly.