Is this 25p or Pal-telecine (2:2) ?

[qwerty1983]

Hello, could you please spare some time to help me with this?

How can i tell the difference between a film broadcasted with 25 progressive fr/s and a film broadcasted with 25 interlaced fr/s (2:2 pulldown) ? In both cases is the film speeded-up (4%)?

2:2 pulldown means that every film frame is transformed in two identical fields? If you mix two identical fields the resulting frame will still have those 'mice-teeth'?


What about this film-sample (only 3 MB --- dvb-t) ?
Is this 25-progressive or 2:2 pulldown? How can i tell the difference?
Please download the sample :
https://mega.co.nz/#!mxYwnTJL!TuRVik5xzlGYdkFa4eoZF0ag9BkRAwmr8q9XfMQNmc4

[Warperus]

Try MediaInfo in text mode.
For your sample it prints the following line:
Scan type: MBAFF
For progressive video it would be:
Scan type: Progressive

Pulldown or not, interlaced video in 4:2:0 subsampling is different from progressive video. Progressive material has common color info for 2x2 blocks in adjacent lines (1+2, 2+3, 4+5 etc.). Interlaced material has common color infor for lines in one field, not the whole picture (1+3, 5+7, 2+4, 6+8 etc.). You can sometimes see this effect in titles and other sharp horizontal edges, one line of the edge.
On the other side, 'mice theeth' is not going to be visible if original source was in progressive scan order. Generally you see them in interlaced material on home video because different fields are separate shots in different times (with 50fps), but if original picture was shot in one time, two fields of one frame reconstruct it without this effect.

In both cases is the film speeded-up (4%)?It depends on what telecine operator did. Generally they are equally sped up.
Amateurs can change fps in different ways - decimate/double some frames, use sophisticated interpolation/morphying etc., but it's not common practice for broadcasts.

[qwerty1983]

Thank you for the reply.

I don't trust mediainfo very much. Is there an empirical (practical?) way to determine if a video is progressive or telecined?

In this particular sample i posted above (broadcasted film - the original was probably 24 fr/s) when i open it with MPC (selecting a bob-deinterlacer from ffdshow's configuration) i notice that every second frame is the same.
When i open the same file in VLC (turning off the deinterlacers) i see no 'mice-teeth' .

In another broadcasted film i can see those 'mice-teeth' and when i'm applying a bob-deinterlacer every second frame is this same. If the two fields are the same why would there be any 'mice-teeth'? In a video that was shot interlaced i can understand it because field1 and field 2 are two different moments. But when you have a 2:2 pulldown (2 identical fields) there should be no 'mice-teeth'. Am i right?


Lots of questions, i'm not the smartest guy around.

[ChiDragon]

What you describe is one proper way to check.

Whether the fields align when they are weaved together for computer display depends on which one is captured first. If they aren't already aligned, all you have to do is remove the first field (or add one to the start if that first field is somehow essential to you). The solution I always use is SeparateFields().Trim(1,0).Weave() but there are other ways to do the same thing.

Thread (http://forum.doom9.org/showthread.php?t=134617)

Warperus has mentioned another issue which is that you need to base the way you handle chroma on whether the encoder used progressive mode or interlaced mode.

[qwerty1983]

Whether the fields align when they are weaved together for computer display depends on which one is captured first. If they aren't already aligned, all you have to do is remove the first field (or add one to the start if that first field is somehow essential to you). The solution I always use is SeparateFields().Trim(1,0).Weave() but there are other ways to do the same thing. Thread (http://forum.doom9.org/showthread.php?t=134617)
Now i get it, thank you. This script 'worked' fine in the broadcasted film with the 'mice-teeth' (2:2 pulldown).
So in this case instead of blending field1a + field1b [coming from Frame 1 of the original film] the broadcaster had blended field1b + field2a [coming from Frame 1 + Frame 2 of the original film]. Am i correct? Is this considered a mistake from the broadcasters' side?

What you describe is one proper way to check.
I opened my sample in avidemux and it says to me that the frame type is I/B/P-TFF so it's telecined (cause when i have a progressive video it says I/B/P-FRM). But again this is based on some digital sign (flag?) that avidemux is able to 'read'.
I still can't find a practical way to distinguise between progressive and telecined(2:2) video . When i apply a bob-deinterlacer (in VLC) on a progressive video it repeats every frame (every second frame is the same) just like on a 2:2 telecined film. The same happens on avisynth. When i have a progressive video and apply a 'separatefields' script it produces two fields out of the progressive frame.

Warperus has mentioned another issue which is that you need to base the way you handle chroma on whether the encoder used progressive mode or interlaced mode.
That's very 'sophisticated' for me to understand right now. I'll leave it for the (near?) future.


So another question i have. It seems that the sample i posted is 2:2 telecined and there are no 'mice-teeth'. Should i apply some IVTC filter? What's the point? It looks like a progressive video. Blending two identical fields produces some artifacts? (what kind of artifacts?)

Sorry for the naive questions.

[TheSkiller]

Warperus has mentioned another issue which is that you need to base the way you handle chroma on whether the encoder used progressive mode or interlaced mode.What makes you and Warperus think so?
I know in some very old threads here on doom9 it is considered correct to use interlaced chroma upsampling for YV12 sources which were enconded as interlaced even if they are progressive (no visible interlacing, the two fields create one progressive frame).

And while that is not "wrong" I'm fairly certain the optimal way is to use progressive upsampling whenever there is no visible interlacing, regardless of how the source's YV12 chroma was downsampled from YUY2 or whatever it was.

I would like to know what some of our chroma experts think about this. Gavino? :)

[ChiDragon]

Well there's this: http://avisynth.org/mediawiki/Sampling

Using point resizing to stress test, interlaced upsampling does look worse on his sample as you say.

So in this case instead of blending field1a + field1b [coming from Frame 1 of the original film] the broadcaster had blended field1b + field2a [coming from Frame 1 + Frame 2 of the original film]. Am i correct? Is this considered a mistake from the broadcasters' side?

Don't use the term blending as that has a whole different meaning. The broadcaster wouldn't consider it a mistake since they are targeting interlaced output.

I opened my sample in avidemux and it says to me that the frame type is I/B/P-TFF so it's telecined (cause when i have a progressive video it says I/B/P-FRM). But again this is based on some digital sign (flag?) that avidemux is able to 'read'.
I still can't find a practical way to distinguise between progressive and telecined(2:2) video .

You aren't really looking for the difference between progressive and 2:2 pulldown (http://en.wikipedia.org/wiki/Telecine#2:2_pulldown), because there is no difference between them when the fields are aligned. The main thing you care about is whether your source shows combing when weaved together for progressive display.

These are the possibilities:

Progressive source, progressive broadcast: should always look progressive when weaved together for computer display
Progressive source, interlaced broadcast: can look either progressive or interlaced when weaved together for computer display depending on which field is captured first
Interlaced source, progressive broadcast: should never happen
Interlaced source, interlaced broadcast: will always look interlaced when weaved together for computer display


Consider this script:
StackHorizontal(Subtitle("Input"),SeparateFields().Trim(1,0).Weave().Subtitle("Phase-shifted"))

Only an interlaced source will show combing in both windows (if the correct field order is determined by the source filter).

But it doesn't tell you whether the chroma has been encoded the interlaced way or the progressive way, which MediaInfo or apparently Avidemux will.

So another question i have. It seems that the sample i posted is 2:2 telecined and there are no 'mice-teeth'. Should i apply some IVTC filter?

No.

[qwerty1983]

Thanks for the answers.

Now i have another broadcasted film (PAL country) with this pattern :

field 1 = FRAME1 of the original film
field 2 = FRAME1 blended(?) with FRAME2 of the original film
field 3 = FRAME2 of the original film
field 4 = FRAME2 blended(?) with FRAME3 of the original film
field 5 = FRAME3 of the original film

Is this another telecining technique?
In this case an ivtc filter will try to de-blend(?) the blended(?) field and then combine(?) the two fields that come from the same original frame?


Progressive source, interlaced broadcast: can look either progressive or interlaced when weaved together for computer display depending on which field is captured first
So if it looks interlaced we can call this material 'field-shifted'? (as i read in this post (http://forum.doom9.org/showthread.php?p=1017936#post1017936) )

[TheSkiller]

Did you step through the video field by field using SeparateFields()? Otherwise you can't be sure if there's blending or not.

Unfortunately, when it comes to broadcasts there is another possibility, a very atrocious one: The film could have been speed up or slowed down by the broadcaster (on top of the PAL speed up/telecine) to better fit their schedules and commercial breaks. Yes, that sounds awful, doesn't it. :rolleyes:
The result is usually a mess if you want to do anything with it.

I've seen it done many times and unless the recording is somehow very important I always discard it.

[ChiDragon]

In this case an ivtc filter will try to de-blend(?) the blended(?) field and then combine(?) the two fields that come from the same original frame?

If you really mean blending (i.e. averaging together frames) then an IVTC filter is not what you need to use. Dealing with field-blended material is not pretty.

So if it looks interlaced we can call this material 'field-shifted'?

Yeah.

[qwerty1983]

Did you step through the video field by field using SeparateFields()? Otherwise you can't be sure if there's blending or not.
Yes, i did it that way.


If you really mean blending (i.e. averaging together frames) then an IVTC filter is not what you need to use. Dealing with field-blended material is not pretty.
I really mean blending. I don't need an IVTC filter then?
Perhaps a filter like srestore or Deblend will do the job? I'll do some reading.

[ChiDragon]

SRestore is the only one I've touched. Unfortunately it couldn't help much with my source.

Basically you can do a high quality bob-deinterlace and live with the blending, or do the same and try to remove it.