Troubles with converting 1080i60 to 720p25 or 720p50

[atl]

Hi.
I have bunch of HD caps made from high-end camera.
Unfortunately camera captures 1080i60.
After some tries didn't got satisfied with results.
What i trying:

Simple Deinterlacing and ConvertFPS(25) does very blurry frames.
With script:
AssumeTFF
separatefields()
selectevery(6,0,1,2,3,4)
weave()
TDeint
Spline64Resize(1280,720)
I managed best results now, but have feeling more can be achieved.

Does somebody have some experience with it?

Edit:

The other way is to interpolate the 60 fields up to 120, then take each 5th to achieve 24p, but also didn't managed to find good motion interpolation plugin, which produce sharp and artefact-free frames

[Adub]

The main reason is that you are going from an NTSC to PAL format. No matter what you do you are going to lose quality, whether its temporal resolution or otherwise.

And do you want 24p or 25p, because you say you want both. For best implementation, you will probably need to use MVTools for interpolation. But I must ask, why do you even need to reduce the framerate? The native format will be fantastically smooth, and is within proper specs, so all common HD hardware should support it.

[Atak_Snajpera]

If I were you I would deinterlace using Yadif to 720p60. You don't have to convert frame rate because every PAL hardware (PS3 for example) supports NTSC framerate as well (23.976 , 59.94 )

[DJ Bobo]

You mean 1080i @ 29.97fps right? There shouldn't be any 60fps camera around shooting at that resolution.
You won't get 25fps progressive from that, at least not clean without blending.
If you want it in PAL format, I would suggest deinterlacing the original input, that gives you 29.97fps progressive. Here you resize. Then you convert to 25fps interlaced:
ConvertFPS(50)
SeparateFields.SelectEvery(4,0,3)
Weave

[Blue_MiSfit]

Do you really need to deliver in PAL? If you're delivering in HD, just deliver 720p60, and all will be fine.

Frame-rate conversion is very evil and should be avoided whenever possible, regardless of how good some methods can be (most of the time)

~MiSfit

[scharfis_brain]

DJ Bobo, you are wrong.
1080i @29.97 (full)fps. is equal to 1080i @ 59.94 (half)fps.
Meaning it has to be deinterlaced to 59.94 full fps.
then one can do a framerate conversion to 50 full fps.

I'd prefer this approach:


xxxsource("1080i60.xxx")
assume?ff()
bicubicresize(1280,height)
yadifmod(mode=1, edeint=nnedi(field=-2))
bicubicresize(width, 720)

super = MSuper(pel=2)
backward_vec = MAnalyse(super, overlap=4, isb = true, search=3)
forward_vec = MAnalyse(super, overlap=4, isb = false, search=3)
MFlowFps(super, backward_vec, forward_vec, num=50, den=1, blend=false)

It gives you pretty high quality deinterlacing at high speed due to splitted downscaling.
And afterwards the framerate is being converted from 59.94 to 50.00 using mvtools2 for high quality

[DJ Bobo]

@ scharfis
There is no full fps and half fps. fps is frames per second. I don't know who came up with this non-sense of calling an interlaced source 50 or 60fps (Wikipedia?)
When he says 1080i60, I understand 60 interlaced frames per second (=120 fields per second), which is obviously not the case.
I don't know how well your script works, but I see a certain nnedi inside it, which means it will be slow as hell, so I wouldn't be very confident about that "high speed" you're talking about.
He's of course free to try it, but I don't think it's practical.
I stay by my original recommendation. I tried it, it looks fantastic on TV (very smooth) -never change a running system ;)-

[EDIT]
On a second thought, I don't know if it's alright to have 720i... if not, then my recommendation only applies to 576i (DVD resolution). 720p50 will be possible if you just use ConvertFPS(50) without the two following lines. Don't know how well that looks and how it compares to sharfis solution with mvtools though.

[scharfis_brain]

whenevery you read 60i or 30i both mean exactly the same thing.
It is an naming issue. I personally prefer 60i, cause it states the nature of the video: 60 fields per second (= interlaced).
Thus you immediatelly know you've got twice the temporal resolution, which you dont have with 30p.
But this doesn't need to be discussed here. It has been discussed several times before and no final conclusion had been drawn...

The is no thing like 720i.

[DJ Bobo]

The is no thing like 720i.

Yeah, I thought so. But then again, there is no 720p25, is there? :confused:

[2Bdecided]

Convertfps is dramatically inferior to mflowfps, except where the latter introduces artefacts.

I don't think the OP has fully explained what they want, or even why they need to convert at all (as others have said, in an HD world, there's less of a reason to convert). There's also a huge difference between 720p25 and 720p50 - which is required? And why not 1080i50?

If it's a "high-end camera" why not put it in the mode that you require? Most/all high-end cameras are switchable. Many notable consumer cameras are not :(.

Cheers,
David.

[atl]

Did to all the footage TDeint(1).AssumeFPS(50).Timestretch, but small portion is fast moving scenes and dancings under with famous piece of music - can't touch the speed/pitch/etc of sound.
For now got best results with:

TDeint(1)
MSU_FRC(2)
Select(5,0)

To Atak_Snajpera:
-Doing this with most of materials

To DJ Bobo:
-1080i60 - exactly 59.94 1920x1080 fields per second
-In HD when FPS is used for interlaced materials the rate of the fields is referd. The standardized rates are p23.976, p24, p25, p30(29.97), p50, i50, p60(59.94), i60(59.94) for 1080. 720 is all of those except the i50 and i60.
-SelectEvery(4,0,3) - breaks the motion and on kind of footage i have looks bad.

To scharfis_brain:
-MFlowFPS generates really weird artefacts on scenes with dancers with fast moving - aka ~1/5 per frame - background.

To 2Bdecided:
-Sadly is already shot, and not in Europe, i just got a bunch of 50Mb/s mpeg2s :(
-Need p24 or p25 or p50 - whatever gives best results.

[2Bdecided]

Has anyone updated the scripts from the alchemist / sal_fps thread to use mvtools version 2 while minimising artefacts?

@atl, can you post a small extract of the most challenging footage, e.g. dancers with fast moving background? It might be possible to tweak various parameters or add some protection so that things fail "gracefully" i.e. going blurred rather than turning into alien shapes!

Obviously with a 60i source, 50p will be more faithful to the original than 25p - it depends what "look" you want. 60i/50i/60p/50p looks like video (very good video, if its HD), 24p/25p looks a bit like film. I'd try both if I was you.

Cheers,
David.

[Mug Funky]

@ DJ Bobo:

the standard notation is that the "i" makes the number after it double.

so 30fps, interlaced is 60i, 25fps interlaced is 50i.

it's written that way on tapes, cameras and decks...

btw, there is a 720p25 and 720p50 on P2 and HDV cameras, and 720p25 setups in the current crop of edit systems.

ugh. too many HD standards, too hard to achieve seamless integration of all of them. why can't all those companies just agree?

[DJ Bobo]

@ Mug Funky
Yeah, I figured that out by searching the internet a bit. It seems like the Blu-ray official document introduced those crappy notations. This is really unheard of, back in the days when DVD was king, everybody was talking about 25 and 29.97fps, be it progressive or interlaced.
Now, they mixed everything up. I'll need some time to get used to this non-sense.

@ atl
Be careful, AssumeFPS accelerates your video. This is acceptable between 24 and 25fps, but between 30 and 50, you've gotta be kidding!
And you did not say why you need 24, 25 or 50. Isn't 29.97fps progressive OK for you? (your only problem seems to be the interlacing).

[atl]

@2Bdecided. There is (c) on it prohibiting me to do this :(.

But the problematic footage was 2 mins - i did it manually. Took me about 2 hours. AssumeFPS(50), then removed ~20% of the frames by hand - mainly in groups by 4-5 before/after scene cuts and got synced audio and 720p50 video.

[2Bdecided]

Blimey - so it was a music video with enough cuts (1 per second) so that you could hide a 20% speed reduction and still stay in-sync?!

That's a radical approach.

Cheers,
David.