Avisynth documentation translation ? & Misc pulldown questions. [Archive]

Wilbert

26th May 2005, 23:43

2:3 or 3:2 pulldown: 23.976 -> 29.97 by adding duplicate fields in a particular way (sometimes falsely called telecine).

inverse pulldown: means undoing the pulldown (sometimes falsely called inverse telecine).

In general, a:b pulldown means that the first frame is represented by "a" fields of video, and the second frame by "b" fields of video, and so on. Thus

field_1 ... field_a field_(a+1) ... field_(a+b)
a ... a b ... b

Pulldown() as AviSynth filter:

DoubleWeave()
Pulldown(0,3)

is the same as undoing a 2:3 (or 3:2) pulldown (ie 29.97 -> 23.976): Suppose you have material on which 2:3 pulldown is applied, ie

AtAb BtBb CtCb DtDb => AtAb BtBb BtCb CtDb DtDb (five frames)

After DoubleWeave:

AtAb BtAb BtBb BtBb BtCb CtCb CtDb DtDb DtDb

Pulldown(0,3) selects frames 0, 3, 5+0, 5+3, 10+0, 10+3, ...

AtAb BtAb BtBb BtBb BtCb CtCb CtDb DtDb DtDb
* * * *

thus getting

AtAb BtBb CtCb DtDb

Removing 3:2 pulldown material goes similar.

I hope it's a bit clear :)

vcmohan

27th May 2005, 03:45

why the name of pulldown? What exactly is pulled down? Then why not pullup for the inverse? The name is not conveying what is being done.

Fizick

27th May 2005, 05:00

Wilbert,
Thanks.
But how understand this: inverse pulldown ... sometimes falsely called inverse telecine.

why "falsely" ?

In Avisynth documentation externalplugins.htm we can read:
IVTC (inverse telecine, aka pulldown removal) filters undo the telecine process

vcmohan,
Yes, i want get some synonyms for "pulldown" word too.

leonid_makarovsky

27th May 2005, 05:18

>2:3 or 3:2 pulldown[/b]: 23.976 -> 29.97 by adding duplicate fields
>in a particular way (sometimes falsely called telecine).

Why is it falsly called telecine? This (http://www.videohelp.com/glossary?T#Telecine) is the definition of TELECINE and it says right there that it the same thing.

>inverse pulldown: means undoing the pulldown (sometimes
>falsely called inverse telecine).

Is there a such thing as inverse pulldown? Again, this is the definition of inverse telecine:
http://www.videohelp.com/glossary?I#Inverse%20Telecine

I didn't see the definition of inverse pulldown. There's a pulldown removal though. So I think the inverse pulldown is a falsy term. :)

--Leonid

Wilbert

27th May 2005, 09:59

I didn't see the definition of inverse pulldown. There's a pulldown removal though.
Yes, you are right. For some reason i couldn't remember that yesterday night.

Why is it falsly called telecine? This (http://www.videohelp.com/glossary?T#Telecine) is the definition of TELECINE and it says right there that it the same thing.
Because originally telecine is the name of the machine that converts film to video:

http://www.screensound.gov.au/glossary.nsf/Pages/Telecine?OpenDocument
http://www.divx.com/support/guides/guide.php?gid=10

But today everyone uses the term also for the process itself.

why the name of pulldown? What exactly is pulled down? Then why not pullup for the inverse? The name is not conveying what is being done.
"The 3:2 pulldown is accomplished in two steps. The first step is to slow down, or "pulldown" the film motion by 0.1%. This speed change is unnoticeable to the viewer, and makes the film travel at 23.976 frame/s.

The second step of the 3:2 pulldown is the 3:2 step. At 23.976 frame/s, there are 4 frames of film for every 5 frames of NTSC video:"

http://en.wikipedia.org/wiki/Telecine

This divx link contradicts this however:

"A telecine machine does something called pulldown, which, in its simplest explanation, "pulls down" an extra frame every fourth frame to make five whole frames instead of four."

I can't find any other link which explain the term pulldown.

Fizick

27th May 2005, 19:11

Thanks, Wilbert.
I think, i understand it now.
I found one more aricle too.
http://www.doom9.org/synch.htm

But there are PAL pulldown too. (24 to 25 conversion),
one frame is added.

But now i have less technical question.
How word "AviSynth" is pronounced (articulated) in right english.
Did anybody hear Ben Rudiak-Gould?
What is transcription?

[a v i 's i n 0]

['ei v i s i n 0]

Wilbert

27th May 2005, 19:25

But now i have less technical question.
How word "AviSynth" is pronounced (articulated) in right english.
We had a thread about this a while ago.

http://forum.doom9.org/showthread.php?s=&threadid=69171

Just pick one :)

Did anybody hear Ben Rudyak-Gould?
No, not that I know.

Guest

28th May 2005, 01:23

Originally posted by Fizick
Did anybody hear Ben Rudyak-Gould?
A God's name should be spelled correctly: Rudiak-Gould.

Wilbert

28th May 2005, 10:43

While we are at it. From http://www.avisynth.org/PullDown

In terms of fields, the 3:2 pulldown sequence is simply "A A B B B C C D D D ...",
where "A" through "D" represent the original film frames. But the capture cards combine
the fields into frames with no respect for the pulldown pattern, and you get this:

A B C D D (30fps)
A B B C D

I don't get this remark (and i don't know who wrote it), perhaps ntsc folks can give some clarification.

I would say that if ntsc people capture from tv/vhs/dvd it would get 29.97 fps (possibly hard telecined). In the example above the capture card appears to receive 23.976 fps and telecines it to 29.97 fps. How and in which situations can that be?

niiyan

28th May 2005, 15:29

Originally posted by Fizick
How word "AviSynth" is pronounced (articulated) in right english.
Did anybody hear Ben Rudyak-Gould?

Stickboy heard how Ben pronounced "AviSynth". See:
http://forum.doom9.org/showthread.php?s=&threadid=76402

Shubin

31st May 2005, 10:56

While translating the Limiter filter from the latest docs, I've found the following problems :

colors were described as
green (pj) green (pj) ... yellow (all) puke (pj), olive (pk)

puke ? Really ?

It also says that parameter values for "show" should be
"show_luma","show_luma_grey" etc, but in fact AviSynth accepts "luma", "luma_grey" - without show_.

Which one is correct ?

Wilbert

7th June 2005, 21:21

colors were described as
green (pj) green (pj) ... yellow (all) puke (pj), olive (pk)

puke ? Really ?
No (at least not that i find), i guess IanB made that one up :) I will replace that one.

It also says that parameter values for "show" should be
"show_luma","show_luma_grey" etc, but in fact AviSynth accepts "luma", "luma_grey" - without show_.
Yes, you should remove show_. I corrected it, thx.

@all,

Please answer:

http://forum.doom9.org/showthread.php?p=661103#post661103

Shall i remove the capture card talk?

Fizick

7th June 2005, 21:59

Yes, you should remove show_. I corrected it, thx.
I am sorry, but I already correct it on CVS (in English too) :)

I wait some comments about 3:2 pulldown too.

IanB

8th June 2005, 03:36

No (at least not that i find), I guess IanB made that one up :) I will replace that one. :o Guilty, "puke" was a descriptive place holder added while bashing the keyboard, I meant to get the colour chart out latter .... Just like the ex bank employee's "Dear Rich Bastard," form letter. :o Please answer: ...
Shall I remove the capture card talk?No. It is trying (badly) to make an important point. Capture cards are the reason this is required.

The reason you need to use DoubleWeave first is that capture cards combine fields in the wrong way. In terms of fields, the 3:2 pulldown sequence is simply "A A B B B C C D D D ...", where "A" through "D" represent the original film frames. But the capture cards combine the fields into frames with no respect for the pulldown pattern, and you get this:

A B C D D (30fps)
A B B C D

In this pattern frame C is never shown by itself. After ....

"the wrong" => "an arbitary"
"combine " => "assembles"
"respect for" => "regard or knowledge of"
"30fps" => "29.97"

The reason you need to use DoubleWeave first is that capture cards combine fields in an arbitary way. In terms of fields, the 3:2 pulldown sequence is simply "A A B B B C C D D D ...", where "A" through "D" represent the original film frames. But the capture cards assembles the fields into frames with no regard or knowledge of the pulldown pattern, and you get this:

A B C D D (29.97fps)
A B B C D

In this pattern frame C is never shown by itself. After ....

Still not great but a little better, maybe someone who is literate can improve it to AAA rating.

IanB

Fizick

8th June 2005, 05:39

About "puke". In russian translation I suggest as "dark olive", but Shubin translate it as "khaki" :)

IanB

8th June 2005, 08:44

Fizick, Wilbert, Others,

"dark olive" and "light khaki" are not bad.

I now remember the agony I had trying to find the name for these 2 colour first time around. ;) For reference :-

The darker "puke" colour is yuv=[ 81, 90, 134], RGB.601=$555600, RGB.709=$565100
The lighter "olive" colour is yuv=[145, 90, 134], RGB.601=$a0a04a, RGB.709=$556b2f

(For the programmers out there Y=81 & 145 are the lowest and highest legal YUV values allowed with [u, v] = [90, 134].)

From color.avsi the 2 nearest colour are :-
global color_darkoliveGreen = $556B2F
global color_yellowgreen = $9ACD32
neither is a very good match.

For the currious here is a scriptpuke=BlankClip(pixel_type="YUY2", color_yuv=(81*256+90)*256+134)
olive=BlankClip(puke, color_yuv=(81*256+90)*256+134)
Interleave(puke, olive)

IanB

Wilbert

8th June 2005, 09:30

Of course we can also add "puke" to color.avsi, that might be the simpliest solution :)

No. It is trying (badly) to make an important point. Capture cards are the reason this is required.

The reason you need to use DoubleWeave first is that capture cards combine fields in the wrong way.
I understand the point being made. What i don't understand is in which cases the cap card is receiving 23.976 fps?

If you cap from tv for example, the cap card will always receive 29.97 fps (60 fields actually).

ac-chan123

8th June 2005, 09:42

But there are PAL pulldown too. (24 to 25 conversion),
one frame is added.

Yes, but most the time it will be only Played faster(speedup).

IanB

8th June 2005, 10:21

Wilbert,

I can see this "puke" colour is going to haunt me. Load and run the script I provided, have a good hard look at the colour, now think about what happens to your carpet when your cat eats something that upsets it's tummy. Yes, yuck! :devil:
I understand the point being made. What i don't understand is in which cases the cap card is receiving 23.976 fps?And the penny finally drops with the 23.976 confusion. What needs to be stated is :-The film goes into the telecine machine at the TV station at 23.976fps, it comes out as 59.94 fields per second, get transmitted, get received, gets decoded as fields and finally reassembled by the capture card as full frames at 29.97 frames per second.

So the Capture card never provide 23.976 only ever 29.97 (well may be an expensive one might ;) )

IanB

Wilbert

8th June 2005, 12:24

That's indeed the logical assumption. But then i don't understand the following:

The reason you need to use DoubleWeave first is that capture cards combine fields in an arbitary way. In terms of fields, the 3:2 pulldown sequence is simply "A A B B B C C D D D ...", where "A" through "D" represent the original film frames.
So after correction, the telecine machine combines the fields in an arbitrary way to get 59.94 fields per second. For some reason i don't believe this :)

I can see this "puke" colour is going to haunt me. Load and run the script I provided, have a good hard look at the colour, now think about what happens to your carpet when your cat eats something that upsets it's tummy

:)

IanB

8th June 2005, 14:28

So after correction, the telecine machine combines the fields in an arbitrary way to get 59.94 fields per second. For some reason i don't believe thisNot quite.

the telecine machine generates fields in a completely deterministic manner to get 59.94 fields per second.

The capture card combines the fields in an arbitrary way to get 29.97 frames per second, with a 5 frame cycle containing 3 "progressive" frames and 2 "interlaced" frames per cycle. We just don't know which are the 2 interlaced frames (yet).

The whole issue is the phase information of the telecine machines internal cycle is not part of the TV signal. The phase has to be determined by examining the resulting decoded and assembled data. And the information needed to infer the phase only shows up when there is sufficient movement in the image to allow detection of interlacing artifacts over any noise present and the phase can change any time there is a cut or other disturbance.

IanB

Wilbert

8th June 2005, 15:21

Perhaps it's much fuss about nothing :)

Suppose the capture card receives the following fields (at 59.94 fields per second):

At Ab Bt Bb Bt Cb Ct Db Dt Db

Usually it combines the fields in the following way

AtAb BtBb BtCb CtDb DtDb

I assume the way the fields are combined is the same for each cap (using the same cap card). Right or wrong?

If you have a crappy card though it will switch the fields:

AbAt BbBt BbCt CbDt DbDt

or worse skip the first one (introducing a phase shift):

AbBt BbBt CbCt DbDt Db ...

Is this what is meant by in an arbitrary way?

IanB

8th June 2005, 16:31

Perhaps it's much fuss about nothing :)Getting it right, concise and understandable will help many.I assume the way the fields are combined is the same for each cap (using the same cap card). Right or wrong?Right for a reasonable card/driver. The original Win98 drivers for my nVidia card randomly switched between TFF and BFF, wrong for a piece of dung.If you have a crappy card though it will switch the fields:

AbAt BbBt BbCt CbDt DbDt

or worse skip the first one (introducing a phase shift):

AbBt BbBt CbCt DbDt Db ...

Is this what is meant by in an arbitrary way?You have got it!

Also drop frame handling screws things as well, my old matrix card drops single fields that makes a nice mess.

IanB

Wilbert

8th June 2005, 16:54

Good :) I will make something nice out of this in the docs.

Fizick

8th June 2005, 20:26

I will make something nice out of this in the docs.
It is really good, but what abot my question? :)
How to translate "Pulldown", what is in other english words,
and why it is named so.

The same question i have about BOB. (I know what is it, but why it is called so).

leonid_makarovsky

8th June 2005, 20:54

terms of fields, the 3:2 pulldown sequence is simply "A A B B B C C D D D ...", where "A" through "D" represent the original film frames. But

This is not 3:2 pulldown, but 2:3 pulldown. I guess it's not a big deal for most of the people, but just wanna make sure you don't confuse the rest of the folx.

3:2 pulldown would be AAABBCCCDD

--Leonid

leonid_makarovsky

8th June 2005, 20:57

How to translate "Pulldown", what is in other english words,
and why it is named so.

Wasn't it answered here:
http://forum.doom9.org/showpost.php?p=660652&postcount=10

--Leonid

Wilbert

8th June 2005, 21:17

How to translate "Pulldown", what is in other english words,
According to Webster it means 'to bring to a lower level'.

I think it refers to the 0.1% slowdown (24 -> 23.976).

The same question i have about BOB. (I know what is it, but why it is called so).
According to Webster it means 'to move up and down in a short quick movement'. In this context the name is a bit misleading because the interpolated fields don't bob, since the top field is moved down one line (after interpolation).

This is not 3:2 pulldown, but 2:3 pulldown.
Correct. I will change that.

Fizick

8th June 2005, 22:43

According to Webster it means 'to bring to a lower level'.

I think it refers to the 0.1% slowdown (24 -> 23.976).

I do not think you are correct here.
I think, this term is from old cinema (projector person pull down (to bottom)the film when film free loop was decreased, and some frames was dropped).

And the are (rare) PAL pulldown, 24 to 25 fps.

So, i wait some comments.

According to Webster it means 'to move up and down in a short quick movement'. In this context the name is a bit misleading because the interpolated fields don't bob, since the top field is moved down one line (after interpolation).

So, there was some old bob-method (without interpolation), with the some quick jumping.
So, Bob-deinterlace - it is "formerly bobbing (jumping)" method...

So, pulldown is filter is not for pulldown (but for removal),
and Bob is not bobed anymore.
:confused:

IanB

9th June 2005, 10:56

@Fizick,

As for "pulldown" the word has become jargin in this industry. I think your basis with old film projectors jumping a sproket hole and the projectionist force pulling the film loop down thru the gate to correct the loop size is very plausible. Also Wilbert's basis with the film slowdown 0.1% most likly has come from this idea.

As for current usage when I see "pulldown" I think of NTSC telecine with it's 3:2 field pattern. Certainly as for it's usage in AVS documents I think this is the intention.

As for "BOB". Again jargin. When an interlaced video stream is field separated and viewed the image wobbles or "BOB's" up and down. This relative motion of the image between top and bottom fields is what "BOB" is referring to. Hence "BOB Deinterlacing" is taking the individual fields and applying compensation for this relative vertical movement. Simply doing the vertical correction is known as "Dumb Bob", adding any motion, interlace or other clever correlation gives a "Smart Bob", very clever correlation gives a "Smarter Bob".

Also in the airline industry BOB is "Babe on board" for cute girl or "Body on board" for cute boy (or more recently "bomb on board" for one very scarred Hostess) :D

IanB

Fizick

20th June 2005, 17:36

Now i found very informative "pulldown" decription ib GreadyHMA plugin:
AutoPullDown - Sets what types of Pulldown processing is desired.
Valid parm values are:

0 - No pulldown, just do pure deinterlace (Force Video). Doesn't drop
any frames so FPS is left at 30 FPS, or wherever. Doesn't decimate
(pentimate?). Use if you have (and want) 30 fps video source or
maybe 25 FPS PAL video.

1 - Auto Pulldown. Automatically decide which frames should be IVTC'd
or deinterlaced, depending upon the settings of the Good and Bad
Pulldown Lvls (see below). Don't decimate. This is best if you
have mixed film and video and wan't 30 fps output. Also better
for PAL sources.

2 - Pulldown only (Force Film). Assume film source, always do IVTC and
never deinterlace. Don't decimate. Use for PAL film source.

Values 3-5 match those above but with frame dropping (Decimation, pentimation?)
Five frames at a time are looked at and the one most likely to be a duplicate
is dropped, thus lowering the FPS from, say, 30 to 24.

3 - Deinterlace (Force Video) but with frame dropping. Use for NTSC video
if you still want to decimate. With video source there really are no
proper fields to drop, but this will drop the ones looking most like
dupes.

4 - Auto Pulldown with frame dropping. The best (most automatic) setting.
Works in most cases, at least if I get all the bugs out.

5 - Pulldown only (Force Film) with frame dropping. This will give the best
results if you have 100% properly mastered NTSC film source with no
video sections and not too many edits. It can adjust for most scene
changes and changes in pulldown cadence, but not mixed up fields.

and below:

The new check box enabled features are:

- Built in 3:2, 2:2, an n:m Pulldown. This does pulldown matching, not removal, so all
frames are displayed.

But next my question is:
Who (or what) is "Greedy"?
:)

Richard Berg

21st June 2005, 16:16

It's computer science slang, not sure of the origin. A "greedy" algorithm in CS is one that chooses the best course of action based only on the current situation, instead of storing or computing a more permanent / global state.

So in video deinterlacing, a greedy algorithm only looks at the area surrounding the current pixel and ignores whatever it might have learned while processing the previous pixel.

Wikipedia entry: http://en.wikipedia.org/wiki/Greedy_algorithm

Fizick

21st June 2005, 22:06

Thanks, Richard,
Tom answer me today with the same info:
In one of my old computer science books a 'greedy' algorithm was a search method that just selected the best available choice over and over again until all selections were satisfied, with no lookahead or backtracking.
So, "greedy" is not a man :)

Wilbert

25th September 2005, 15:22

@IanB,

If you have a crappy card though it will switch the fields:

AbAt BbBt BbCt CbDt DbDt

Is this what is meant by in an arbitrary way?
You have got it!

It's time to resurrect this old thread, because it's still not clear to me :)

In which sense is this actually true? I mean does it hold for capped avi with swapped fields? In that case the problem isn't corrected by using DoubleWeave (after all how should avs know the fields are swapped).

Does it hold for capped mpeg-2 where by error the field-dominance is switched? Suppose you have telecined material with layout:
AtAb BtBb BtCb CtDb DtDb

Consider the field-dominance:
BFF: Ab At Bb Bt Bb Ct Cb Dt Db Dt (1)
TFF: At Ab Bt Bb Bt Cb Ct Db Dt Db (2)

Let's say (2) is correct, and (1) is caused by stupid capture card (and thus should be corrected).

Applying DoubleWeave you get:

(1) AbAt BbAt BbBt CbBt CbBt DbBt DbCt DbCt DbDt
(2) AtAb BtAb BtBb BtBb BtCb CtCb CtDb DtDb DtDb

So, how does DoubleWeave correct the wrong field-dominance? :confused:

IanB

27th September 2005, 02:33

@Wilbert,

On re-reading I think I see where some of the confusion is coming from, there are 2 issues here being mixed together.

1. Capture card drivers that assemble fields into frames the wrong way up, i.e. use a top field for the bottom and a bottom field for the top. Here there is clearly a right and wrong way.

2. The way a capture drive groups fields together, i.e. is the top field temporally before (TFF) or after (BFF) the bottom field. Here there is no right or wrong way it is purely a reference point perspective issue.# Exchange fields and choose grouping
SeparateFields()
ComplimentParity()
DoubleWeave()
SelectEven() # Choose original field grouping
# SelectOdd() # Choose phase shifted field grouping

In which sense is this actually true? I mean does it hold for capped avi with swapped fields? In that case the problem isn't corrected by using DoubleWeave (after all how should avs know the fields are swapped).AVS does not know, the user has to know and write code to fix it. Remember swapped fields are not the same as phase shifted assembly.Does it hold for capped mpeg-2 where by error the field-dominance is switched?But field-dominance is not quite exactly the same as field order. For live video there is no dominance, fields are purely sequential each snapping the world each 60th of a second. Only when paired together in an AVI does TFF or BFF come out to play. For Telecined film material there is also a phase relationship which relects in field-dominance.Suppose you have telecined material with layout:
AtAb BtBb BtCb CtDb DtDbAnd even here there is a true transmission temporal order to the fields. The difference from live video is there is also an aditional temporal grouping from the source images, 24fps film.Consider the field-dominance:I think you really mean "Assumed assembled field order"
BFF: Ab At Bb Bt Bb Ct Cb Dt Db Dt (1)
TFF: At Ab Bt Bb Bt Cb Ct Db Dt Db (2)

Let's say (2) is correct, and (1) is caused by stupid capture card (and thus should be corrected).But your rendering of (1) is incorrect, it would be
Ab At Bb Bt Cb Bt Db Ct Db Dt (1)Applying DoubleWeave you get:

(1) AbAt BbAt BbBt CbBt CbBt DbBt DbCt DbCt DbDt
(2) AtAb BtAb BtBb BtBb BtCb CtCb CtDb DtDb DtDb

So, how does DoubleWeave correct the wrong field-dominance? :confused:Simple answer is it doesn't (directly). It provides access to the phase shifted variant of the field grouping. To exchange fields you need to ComplementParity() somewhere.

IanB

Wilbert

28th September 2005, 15:03

@Wilbert,

On re-reading I think I see where some of the confusion is coming from, there are 2 issues here being mixed together.
It's not confusion, i understand the difference between the two. I just want to know which of the two situations the original docs were refering to.

1. Capture card drivers that assemble fields into frames the wrong way up, i.e. use a top field for the bottom and a bottom field for the top. Here there is clearly a right and wrong way.
Ok. But DoubleWeave doesn't correct this. So, the docs were not refering to this case, right?

2. The way a capture drive groups fields together, i.e. is the top field temporally before (TFF) or after (BFF) the bottom field. Here there is no right or wrong way it is purely a reference point perspective issue.
If you cap to avi it doesn't matter, since you need to find the TFF/BFF info manually.

But what about mpeg-2? I assume it can happen that the TFF flag is added to all fields, but they are actually BFF. Of course this matters if you want to apply temporal smoothers, deinterlace/ivtc or whatever. Are the docs refering to this case?

IanB

30th September 2005, 17:02

It's not confusion, i understand the difference between the two. I just want to know which of the two situations the original docs were refering to.Okay, not confusing but there is some ambigulty. Looks to be the phase shift case to me, but it certainly could be expressed clearer.

Ok. But DoubleWeave doesn't correct this. So, the docs were not refering to this case, right?But DoubleWeave() is the correct tool to fix this problem, I think the example was incorrect. Maybe the code to fix swaped fields (in post above) should be included as a separate section.

But what about mpeg-2? I assume it can happen that the TFF flag is added to all fields, but they are actually BFF. Of course this matters if you want to apply temporal smoothers, deinterlace/ivtc or whatever. Are the docs refering to this case?Pass. Perhaps Mr Graft has an opinion on this, he is the DGDecode expert.

IanB