Why is 24 sometimes interlaced other times not.

[Wam7]

I'm capturing episodes of 24 from BBC2 using:

VDub Sync 1.4.9.2 with VFW drivers
VBLE beta or Morgan 3009 at 98%
PAL I
768x576
all others at default

At first my captures were interlaced which was what I expected but then on other occasions I've noticed that they weren't interlaced, which is what I prefer as it gives better xvid encodes.

I actually think it's something I'm doing differently rather than the broadcast but I can't figure what it is. The only thing I'm leaning towards is that I've been using Enable DirectDraw Aceleration Both Fields which Deinterlaces it some how for that particular program.

[neuron2]

It's PAL so you can have simple field shifting that can be undone easily using Telecide or any other decent field-aligning filter. Always try that before concluding that the source is truly interlaced.

[Wam7]

Thanks, I've been using TomsMoComp in Avisynth to deal with the interlacing. I'm just seeing what Telecide filters there are and if they'd do a simpler/better job than TomMoComp.

I'm still bemused as why it sometimes doesn't have that interlaced look.

[Swan]

@Wam7

I'm still bemused as why it sometimes doesn't have that interlaced look.
It's quite simple to explain. "24" is the US series, right?
It's most likely shot on film (24 fps), as are most US series, where the producer has export of the series in mind.
When the BBC buy the program from the US producer, they get or make a PAL 25 fps copy from the film material. And that conversion, from film (24 fps progressive) to PAL (25 fps interlaced) is done the same way as films are transferred to DVD (the film is sped up to 25 fps, one film frame is used to create the 2 fields in an interlaced frame and the audio is sped up to match). This looks progressive when you capture it. No "mice teeth" because each frame is made up of two fields *emanating from one and the same film frame*. So both fields in one frame contains the same picture info.

Now, perhaps the BBC or whoever they hired sometimes uses a different method to transfer the film material to broadcast-ready PAL?
This could explain why "24" looks interlaced when you watch it on your computer monitor. It could be field-shift, as Neuron2 suggests, but it could also be that sometimes what you see is a PAL copy of "24" that's been made from the NTSC copy, not the film, which is the correct way. Going from Film to NTSC then to PAL is what gives awful picture quality and what's known as "blended fields".
But I doubt the BBC would buy such terribly transferred material and broadcast it.

If you can see BBC Prime, check out the old comedy series, like Harry Enfield and Chums, which is on late Fridays. If you capture it, you will see a mix of interlaced material (shot with a video camera) with mice teeth and non-interlaced-looking scenes (shot on film, transferred to PAL using each film frame to create the two fields in an interlaced frame).

[Wam7]

Thanks for the extremely informative info. I follow what you are saying but I surmise that the BBC broadcast in exactly the same way each time. Therefore it should look exactly the same way each time. Last week I captured it and I did not have to use TomsMoComp at all whereas this week it had the 'mice teeth' look to certain frames in moving areas.

Now if you are saying that the BBC may indeed output it differently each week then that clears up my conundrum. If I can totally rule out anything I'm doing then it's down to the broadcast.

Just to give you all details. I'm using Telewest Digital Cable as a source via a S-Video lead into the composite input of my WinTV card. The only other thing I have acitvated is Squish Luminance range.

PS. I did try the Telecide and it worked fine, although I prefer the look of TomsMoComp(1,5,1).

[scharfis_brain]

try

telecide(post=false)

is it still interlaced?

[Wam7]

@scharfis_brain
Yes I did try it and it still gives that 'mice teeth' look in certain frames. Telecide() without any parameters does get rid of it.

[jggimi]

You might check out this tutorial (http://www.doom9.org/ivtc-tut.htm) on Inverse Telecine processing. Hakko wrote a terrific section on film and NTSC transfer to PAL. He puts PAL transfers into three categories, with suggestions for handling each one. He calls them the Good, the Bad, and the Ugly. You'll get suggestions for frame analysis, and recommended filter settings to correct each type.

[Wam7]

Thanks. I did capture 24 again last night and this time there were no horizontal lines so no need for Telecide/Deinterlacing. I read through the whole of that tutorial but didn't gleen any information that would explain as to why the change.

[dar1us]

Information:

24 Is recorded to FILM, so is at 24fps. The version sold to the UK is the 24fps FILM format. This is then sped up to 25fps and should still be found progressive.

BBC is odd as they broadcast it interlaced sometimes?! I Have no idea why. As far as I have seen, MOST stuff, films, programms, the news are progressive (progressive Video camera's). Most Dramas are progressive.

The fact that they broadcast it progressive doesn't matter to them as it doesn't concern most people. The fact of the matter is that you are not supposed to be capturing it, but that isn't a topic for here.

Telecide() usually gives the best result with anything captured VIA S-Video in the uk, post=false is always a good start as the postprocessing looses details. Try using postprocessing with telecide then playing around with the thresholds and other settings until you get frames that show NO combing artifacts. Try and tweak your usage of telecide (thresholds) so they are just beyond the combing barrier.


harrison

[dar1us]

I will capture 24 from SKY Digital on BBC 2 on sunday night. If you do the same, we can compare results. This will be the new episode, forget the repeats. This way, we can compare individual frames.

If there is anyone else out there, that would be great to see what your system makes of it.


harrison

[Wam7]

Ah excellent. That's the same conclusion I was coming to after the reading I've been doing over the past few days but you have a greater breadth of knowledge. ;)

On the Sunday just gone, I played around with all permutations of my capturing just to see if I could get it to show combing artefacts as it has in the past from time to time but to no avail. This lead me to the conclusion that it's the BBC broadcast and not some gremlin setting in my capture process.

3 meg Sample Clip (http://www.wmarriott.pwp.blueyonder.co.uk/24Clip.avi)

This is a capture & I encode I did without the need for any Telecide or TomsMoComp kind of filter.

I will be capturing on Sunday night as usual (in spite of the BBC's wishes ;)) from Telewest Digital Cable. I capture the one on BBC2 first then switch to BBC3 for next weeks but they will be different files. We will see how this one comes out. Look forward to comparing!

[XenoDaSouljah]

What is mice teeth ? please explain :)

[scharfis_brain]

"miceteeth" is another word for interlacing oder combing.

[dar1us]

Okey - I looked at the sample. The quality was quite good,

OT on Sample: I would reccomend that you dont resize so much, you are actually resizing OVER the broadcast res. It is broadcast at 544x576 I believe. Possibly 720 at the MOST.

I will capture it tonite, and look forward to comparing.


harrison

[dar1us]

Ok - all capturéd, the capturé went smoothly on the outsét (I was using a néw codec). But on playback, the quality créated with this lossless codéc (naming no namés) seems very poor indeed. Like it is adding noise itself to compénsaté for blocks that it makés.

I usually usé HuffYUV or XVID. I know I should have playéd around with it beforé, but... Grrr.

I will turn it yét into a 1/2 CD XVID for you. I like the resoloution 576,320 personally. I missed the first 5 seconds of the 24 logo though. This is cos I stopped the capture BECAUSE I saw it was capping (as default suggested) 480x576 for SVCD use.

I will get on that soon, just encoding the Andrew Marr sketch on Dead Ringers, god that was funny. :-

*Comedy Removed*


harrison

[Wam7]

Yep got it as well. Sorry for the delay but got no email notification! PM me when you're ready.

With regards to the resolution of the capture, are you saying I would be better off at least dropping the capture resolution to 720x576 instead of leaving it at 768? I don't actually do any resizing just capture at full D1 and crop the black borders. I did try resizing (after capture) to 640x??? but didn't notice any improvement in quality and there was a very slight loss in detail. (plus had problems with B-frames and file size as they need to be 350meg!)

I can see the logic behind this especially if the broadcast output is 544 or 720 but I have no way of knowing what resolution a program is output at and would be interested as to how you can find this or make an educated guess.

PS. I will want that Dead Ringers from you as well! :) (used to listen to it on the Radio... eye wateringly funny!)

[dar1us]

Finding out resoloution. Clever way of doing it. I worked this out myself as far as I can remember. You record a VERY actioney scene, then count how many compression blocks you can see. Each block is 16x16, so if you counf 10, then it has 160 horizontal res:)

I will capture it again this week (24) because quality is apauling!


harrison

[scharfis_brain]

ahem...but full-D1-Res. is not suitable for DivX, because it is not 4:3!
for DivX you need square pixels, 768x576, 640x480 or others.
720x576 will give you a wrong Aspect Ratio

[bachopin]

Hi,

I just came across a PAL TV captured material and I suspect
that it belongs to the 'blended fields' category you mentioned here:

"Going from Film to NTSC then to PAL is what gives awful picture quality and what's known as "blended fields"."

When I did a separatefields() on the captured material I noticed
the individual fields have blended effects. Simply I didn't undesrtand
how that might have happened. Just after I read your lines I realized
that you shed some light on my problem :-)

So is it imaginable that a TV Brodcasting corporation could
use such a terrible source material ?
I've never dreamed they would do.

I think there isn't any remedy for these type of footage, there is ?

[Kika]

Oh, there are correct Conversion from Film -> NTSC -> PAL, but unfortunatly, they are rare...

The Blending Effects are from a Framerate-Conversion 29.97 -> 25.

[dar1us]

Where as they should IVTC then speed up to 25 and pitch shitf.

I started a thread about this kind of material, repal (i think) should fix it though I have had no luck, Repal is an AVIsynth plugin.


harrison

[Swan]

@ bachopin
I just replied to the PM you sent then checked out this forum only to see that you had posted in this thread and with almost the exact same phrasing as your PM to me. In the future, please post in a thread *or* PM, but don't do both.

Anyway, blended fields occur in two scenarios.
If the source material was shot on NTSC and then converted to PAL, blended fields is a side effect that can't be avoided.
Converting NTSC (29.97 frames/59.94 fields per second) to PAL (25 frames /50 fields per second) naturally gives blended fields, as Kika pointed out.
The second way to get blended fields, which, unlike the previous example can be avoided, is the film-->NTSC-->PAL conversion, instead of film -->PAL mentioned earlier.

No, there is no way to "fix" it, ie.e bring back progressive frames from a bad film-->NTSC-->PAL conversion. The material is ruined, you can't recover frames and fields that were lost in the conversion and no longer exist in the material. But for encoding puposes, it's better to leave the blended fields than deinterlacing, it will just ruin the material even more. For SVCD and DVD encoding, just leave the blended fields and do not deinterlace.

Like I wrote in the PM, I see blended fields a lot on VH1 and MTV Europe (MTV Nordic). They play loads of videos that have been converted either from NTSC to PAL or have been converted from NTSC, instead of the film source the NTSC copy was made from.
The more high quality the station, the less risk of blended fields from bad conversions. I don't see it on the national TV station or other stations that are "public service" (not commercial, but sponsored with tax money = wealthy stations). This is also the reason why I doubt the BBC2 would show "24" that has been converted to PAL the "el-crappo way". :)

/Swan

[bachopin]

@ Swan
>In the future, please post in a thread *or* PM, but don't do both.
sure, sorry I made a mistake by not doing that in the first place. :(

In regarding to blended fields, you mentioned that:
"If the source material was shot on NTSC and then converted to PAL, blended fields is a side effect that can't be avoided."

The footage in question was originally a 24fps movie not NTSC video
and I wonder if the effect could have been avoided by correct IVTC where the process deals with fields not frames as many 3rd party plugin can do correct IVTC - sometimes it needs some manual tuning but most of the time it goes automatically.
I believe during the process they simply left out one frame out of every 5 frames in order to get 24 fps and speeded up to 25 fps.
So in a nuthsell the result the same as if they converted from original NTSC footage.

I hope you understand what I tried to say.
Thanks again

Best regards

[Kika]

Like Swan wrote: There are different ways to convert from NTSC to PAL, even the source was Film.
The "clean" way is to do a PALSpeedup (23.976 played with 25 FpS).
What you get is a clean, progressive Video.

The bad way is to do a framerate conversion. This can be done in two ways:

1. Convert from 23.976 to 25 by filling in extra Fields.
Such source can be restored to progressive, but it's not easy to do.

2. Convert from 29.97 to 25. That's the same methode to convert footage shot in interlaced NTSC. Using this way, a couple of fields will be blendet with other Fields/Frames.
Here, some Fields/Frames are destroyed during the process. There is no way to get them back. Best thing to do with such source is to keep it interlaced.

[Xesdeeni]

dar1us wrote:24 Is recorded to FILM, so is at 24fps. The version sold to the UK is the 24fps FILM format. This is then sped up to 25fps and should still be found progressive.Actually, 24 is a hybrid. I've taken a close look at the 24 DVDs, and although the content was captured on film, it was edited and post-processed in video format. Since FOX is using anamorphic 480p for their DTV format (720p HDTV coming in 2004), I assume that is the format in which it was edited (I don't have an HDTV at this point). But regardless, it was edited and post-processed in 480p at 60fps, or 480i at 30fps, because the post-processing effects (boxes moving around) differ on every field (of the 480i DVDs). The content itself however, was definitely captured on film, because the 3:2 pulldown is evident as well.

So when converted to PAL, they could have done a conversion without speed-up (see http://www.geocities.com/xesdeeni2001/StandardsConversion), they could have converted it to 24 fps and then sped it up, they could have done a combination, or they could even have re-edited it for PAL (doubtful).

Xesdeeni

[dar1us]

I am really sorry to fog you off again WAM7 with the capture comparison, but I got a better resaon this time than a poor codec used.

To capture, I have to lug my beast computer downstairs. This is posing a problem at the moment as yesterday I came off my bike and fractured my elbow, twisted my knee round, dislocated my arm/shoulder (hasn't been reset yet) and tore up my side as I was cycling with shorts and no shirt:). It really hurt (miles is going for the sympathy vote here:)) and wont be back on the saddle (not double entendre) for a few weeks:( which is my second fave sport, after skiing of course:). It's is a shame with the trans-european DECENT weather.

Maybe the next week, or I could persuade my brother to take it downstairs for me, but I don't trust him, I know he will drop it.

Good luck with your capture regardless.

Would hybrid make sense that some episodes are interlaced and others are not? I mean, not just sections, with most tv dramas/comedies made in the UK, when on DVD, they are progressive with interlaced credits (a video layer on progressive material?), ie Spaced or Shackleton (just watched both recently). Would this make sense to explain why such material is interlaced not all the way through, though saying that, this isn't the problem that WAM7 is expressing is it?


harrison


harrison