Framerate conversion thread 2012

[hanfrunz]

Hello everyone,

i found some "old" threads about framerate conversion.
Is someone using avisynth for NTSC<->PAL (interlaced) conversions in a professional enviroment? Are there new modern "2012" tools (plugins/scripts) that can be used?

regards,
hanfrunz

[nhope]

This is what I do for interlaced>interlaced and the results are great.

PAL to NTSC DV

AviSource("whatever.avi")
ConvertToYUY2(interlaced=true)

QTGMC( Preset="Slower", SubPel=2, Sourcematch=3, Lossless=2 )
super = MSuper(levels=1, pel=2)
MFlowFps(super, QTGMC_bVec1, QTGMC_fVec1, num=60000, den=1001)
Spline36Resize(720,480)

AssumeBFF()
SeparateFields()
SelectEvery(4,0,3)
weave()

NTSC to PAL DV

AviSource("whatever.avi")
ConvertToYUY2(interlaced=true)

QTGMC( Preset="Slower", SubPel=2, Sourcematch=3, Lossless=2 )
super = MSuper(levels=1, pel=2)
MFlowFps(super, QTGMC_bVec1, QTGMC_fVec1, num=50, den=1)
Spline36Resize(720,576)

AssumeBFF()
SeparateFields()
SelectEvery(4,0,3)
weave()

I convert to YUY2 because I get this problem if I weave YV12.

(edit: there are more variants of these scripts here, including HD versions)

[TheSkiller]

Originally Posted by nhope

I convert to YUY2 because I get this problem if I weave YV12.

This is just because of the way you view the video, it is not actually broken.
If you have an interlaced YV12 video and watch a frame of it on a PC monitor using VirtualDub or the like the color space needs to be upsampled to RGB somewhere. And that's where the problem occurs. If you do not convert – just for displaying, not for encoding – to RGB within AviSynth using ConvertToRGB(interlaced=true) you will see the chroma getting mixed between the two fields of each frame, giving this "chroma-ghost".

One of the reasons I always recommend to convert to RGB at the very end of a script while checking things. For encoding of course it should be removed.

[nhope]

Thanks very much TheSkiller. This makes sense and I will check it out.

[wonkey_monkey]

The Wikipedia page on standards conversion talks about "phase correlation" and makes it out to be the best method - are any of the existing AviSynth solutions anything like that?

David

[Didée]

Diggin' an old post about phase correlation: <click>

Briefly: Phase correlation tells you that there IS something. But it doesn't tell you WHERE it is.

Practical examples: straight camera panning -> phase correlation works beautifully. Closeup of a bustling anthill -> phase correlation is utterly useless.

[Mole]

What about keeping the same frame rate, but using pulldown?

Obviously this would only work for PAL->NTSC, but not NTSC->PAL.

[manolito]

Of course, have a look at DGPulldown. Just keep in mind that the source has to be progressive...


Cheers
manolito

[Mole]

Yeah, I forgot. Pulldown works only if it's progressive.

[manolito]

Anyone tested the MSU Advanced Frame Rate Converter?


Cheers
manolito

[-Vit-]

I tried MSU quite a while ago. Visibly very blocky where motion is complex - at the time I much preferred the how MFlowFPS dealt with difficult areas (that blurry wobbliness). Of course we now have SVP/Interframe improving things in that area. Dunno if MSU has been updated more recently

[Mug Funky]

i'd vote SVP, or super-duper-wrapper scripts that incorporate it as a motion engine.

it's a little less prone to failblocks and faster than mvtools (though based on it), and allows us to use the GPU for fancy things.

currently i'm using it to go from sharp 720p60 to motion-blurred 720p24 and its very rarely gives artefacts, and when it fails it just gives a frame-blend rather than a warpy-thing. i'm about to build a proper converter with it in the coming weeks, so i'll post back here with results