QTGMC Deinterlacing Script (v3.32)

[-Vit-]

Originally Posted by ajp_anton

Is it possible to run this on YV16 and YV24?

Convert YV16 to/from YUY2 (lossless) and QTGMC will process it natively. You need SSE2Tools and (from the first post): "For YUY2 input you must use mt_masktools-25.dll. Although not strictly 'correct' mt_masktools-25.dll can be used with SEt's Avisynth 2.6MT for multi-threaded usage."

I would guess you could force YV24 through there with various splitting and merging, but I've never tried it.

[Boulder]

Originally Posted by chainik_svp

-Vit-
a major problem with HD is that we cannot run enough threads to use up the CPU before running out of memory

It was the first real problem we faced with SVP The current solution (modified Avisynth version) is good for SVP, may be it can be good for you too.

Can SetMTMode be used with your Avisynth build or is it natively multithreaded?

[chainik_svp]

It's the "official" 2.5.8 MT except rewritten distributor and modified CacheMT code.

[Boulder]

Ok, thanks I think I'll try it since I use the various MVTools2 stuff a lot (mostly QTGMC and a custom denoiser).

[zerowalker]

Originally Posted by yup

zerowalker!
try

SetmtMode(3,4)
AVISource("tape1.avi").Trim(0,129057)
AssumeTFF()
SetmtMode(2,4)
source=ConvertToYV12(interlaced=true).Crop(16,12,-16,-12)
bob=source.nnedi3(field=-2,threads=1)
super = bob.MSuper(pel=2, sharp=1)
backward_vec2 = MAnalyse(super, isb = true, delta = 4, blksize=16, overlap=8, chroma=false)
backward_vec1 = MAnalyse(super, isb = true, delta = 2, blksize=16, overlap=8, chroma=false)
forward_vec1 = MAnalyse(super, isb = false, delta = 2, blksize=16, overlap=8, chroma=false)
forward_vec2 = MAnalyse(super, isb = false, delta = 4, blksize=16, overlap=8, chroma=false)
bc2 = bob.MCompensate(super, backward_vec2)
bc1 = bob.MCompensate(super, backward_vec1)
fc1 = bob.MCompensate(super, forward_vec1)
fc2 = bob.MCompensate(super, forward_vec2)
Interleave(bc2,bc1,bob,fc1,fc2)
dfttest(sigma=100.0,tbsize=5,sbsize=16,sosize=12,threads=1)
Edibob=SelectEvery(5,2).nnedi3_rpow2(2,nsize=4,cshift="Spline36Resize",fwidth=720,fheight=576,threads=1)
Edibob.SeparateFields().SelectEvery(4,0,3).Weave().AssumeTFF()
QTGMC( Preset="fast", FPSDivisor=2, Ediext=Edibob,EdiThreads=1 )

Tune sigma, 100 good starting point. Script include work with borders.
-Vit-
This script better denoise than QTGMC. I remember Your suggestion, but for analog capture need external prefilter.
Also filtering on bobbed give better result than on fields.
Work slow 1.5-2 fps on Quad Core.
yup.

Is this considered for PAL?
As there is quite some difference in denoising and restoring PAL compared to NTSC (PAL is much worse sadly).

As you will first have to blur every other Chroma line, as PAL is "YV12" but in disguise as YUY2 (or something like that).

Cause if that´s considered, i will gladly try it out and compare it against the one i tried before (the lossless sourcematch thing).

EDIT:

needed to add

Mergechroma(SeparateFields().Blur(0,1).Weave())
ChromaShift(-2)

to make it more PAL friendly (ChromaShift depends on source etc i guess).

And it seems pretty neat, though 100 sigma was way to smudge looking for me, around 25 seems nice, but i would like to try to keep a bit more sharp edges if possible, as this does very good on the noise,
but it blurs a bit much on some details.

Noticed that my source is 50 fps, as it was much smoother without selecteven().
If i want to burn it on a DVD, can i just skip the QTGMC in the end, and use 50i flag on the mpeg2 file?

[nhope]

-Vit-, I have tracked the cause of this error to the modified eedi3.dll from your plugin pack using the updated Avisynth 2.6 interface, linked to in the first post. I replaced it with the eedi3.dll from your previous modded plugins package and all seems OK. Note I am on Windows XP x64 SP2.

What really threw me in tracking down the error was that even when I had renamed that file to eedi3.dll-VIT-NEW-INTERFACE it still threw the error. I often rename old or alternative dll files by appending the file extension like that, and I had no idea that AviSynth would still somehow read them. I have many other ones renamed in a similar way that appear not to be causing any problems. Perhaps I should keep those sort of files in a different folder.

[-Vit-]

Always read the fine print... From the first post:

Originally Posted by -Vit-

If you use EEDI3 you may need the Visual C++ 2010 Redistributable

tritical uses the OpenMP library in EEDI3 (only). It creates a dependency (VCOMP100.dll) cannot be safely removed with static linking. Edit: The previous EEDI3 was tritical's original (I did not add the mod to that one), so it used VCOMP.dll from Visual Studio 2005. You already had that dll. I used Visual Studio 2010 to compile the 2.6 version and that needs the newer VCOMP100.dll

[Gavino]

Originally Posted by nhope

What really threw me in tracking down the error was that even when I had renamed that file to eedi3.dll-VIT-NEW-INTERFACE it still threw the error. I often rename old or alternative dll files by appending the file extension like that, and I had no idea that AviSynth would still somehow read them.

Rename to something like eedi3.dll.VIT-NEW-INTERFACE (ie replace first '-' by '.').
See this post.

[nhope]

Thank you -Vit- and Gavino. During my troubleshooting, because of what Google turned up, I did actually install the Visual C++ 2010 Redistibutable. But as I'm on XP x64 I installed the x64 version, and it didn't help. I guess I need to install the x86 version you link to?

By the way, is there any reason you're holding the thread title and links in the initial post at 3.32 and not 3.33? It's easy to miss your 3.33 post, as I did until yesterday.

[-Vit-]

Originally Posted by nhope

I guess I need to install the x86 version you link to?

Yup. Avisynth is 32-bit. Redistributables should match the application not the OS.

Originally Posted by nhope

By the way, is there any reason you're holding the thread title and links in the initial post at 3.32 and not 3.33? It's easy to miss your 3.33 post, as I did until yesterday.

Just keeping everyone on their toes... No, I didn't really want to release 3.33 because I haven't tested it enough (who likes support queries?) and it doesn't add much. But it suited an earlier problem, so I let it out gently.

[zerowalker]

How can i disable everything in QTGMC except the part that makes shimmering alot smoother?
I am doing a 25 fps (50i) video by filtering with dfttest, then going to qtgmc for the anti shimmering and then reinterlacing.

I think the way sound pretty bad, so if i am doing something stupid here please tell me.
I need to keep it interlaced as DVD only support 50i so can´t keep the 50p qtgmc gives.

[AzraelNewtype]

Originally Posted by zerowalker

How can i disable everything in QTGMC except the part that makes shimmering alot smoother?
I am doing a 25 fps (50i) video by filtering with dfttest, then going to qtgmc for the anti shimmering and then reinterlacing.

I think the way sound pretty bad, so if i am doing something stupid here please tell me.
I need to keep it interlaced as DVD only support 50i so can´t keep the 50p qtgmc gives.

This is an ill-thought out idea and will accomplish nothing. Just separate fields, filter, and then weave.

[Boulder]

Not that simple if you use temporal filters

[kolak]

Found some issue- DTC=5 causes problems on scene changes- creates frames with mixed fields.
In the same time it helps a lot with flickering on static text.

[-Vit-]

May need to tweak thSCD1 (scene change detection setting, same as MVTools setting of same name) for different DCT settings.

[kolak]

That's what I thought

[Mole]

When using QTGMC(InputType=1), I kinda like the way it smooths out noise such as CCD noise.

What kind of denoiser is it using, and would I be able to use only that particular denoiser in order to get the same effect?

The source is of course progressive.

[Betsy25]

Although having followed the instructions, simply indexing the .VOB files in MeGUI produces the following error screen.
Win7 32bit, Avisynth 2.5

[-Vit-]

Mole: QTGMC uses denoisers as an integral part of the algorithm, not specifically to denoise, but it does have some denoising effect. It uses MDegrain/TemporalSmooth in three stages:
0. A binomially-weighted combination of TemporalSmooths to remove shimmer from the source for motion analysis - this denoise doesn't directly appear in the output.
1. Do the same again, but this time with motion-compensated MDegrains, to produce initial deshimmered output (and to get some super-resolution-like detail-borrowing from neighbor frames)
2. A single MDegrain at the end for a final calming

So it's not just a simple call to a denoiser as you might have thought, it's the core algorithm you need. When you set InputType=1 you switch off all the deinterlacing and you are left with that algorithm alone essentially. I guess there are a few tweaks:

You can can skip stage 0 if your source (progressive, InputType=1) has no shimmer by setting TR0=0 (and some other related settings):
QTGMC(InputType=1,TR0=0,Rep0=0,SrchClipPP=2)

There is some extra processing between stage 1 and 2 - resharpening/repair steps to reduce blur from the MDegrain. If you *really* want the MDegrain only, then:
QTGMC(InputType=1,TR0=0,Rep0=0,SrchClipPP=2, Sharpness=0, Sbb=0,Rep2=0)

You may not want that, it will lose a little detail. Maybe best just to tweak sharpness:
QTGMC(InputType=1,TR0=0,Rep0=0,SrchClipPP=2, Sharpness=0.2)

Source match is a rather strange, overkill setting for progressive but it's possible - will retain the most detail but will only impact the noise very slightly too. I don't exactly recommend it, but you could try it:
QTGMC(InputType=1,TR0=0,Rep0=0,SrchClipPP=2, SourceMatch=2) # Maybe add very low Sharpness too

___

Betsy25: If it's the indexing stage, then maybe it's nothing to do with QTGMC. That missing dll is from Visual Studio 2003, so it's an older app or plugin causing the problem. Not sure where you can get the dll, I don't think it's in any Microsoft redistributable. Could just search for a version, but be careful...

Some of the original versions of the plugins for QTGMC do rely on that dll. The Modded plugin pack from the first post removed the need for that dll from everything except SSE2Tools (for YUY2 input). Maybe try getting that pack first and see if it helps.

[Mole]

Thanks a bunch for the tips. I've been using QTGMC(InputType=1) with progressive source that are very very grainy such as from many BDs. I of course resize them to whatever resolution I need first (ideally I'd rather like to use it at 1080p, but it'd just be way to slow if I run neat afterwards), I then use QTGMC on it to get rid of the most nasty grains, then when I use Neat Video after, it looks quite good. It's then sufficient to use Neat with minimum settings. Usually on very grainy sources, I had to crank up Neat, but the problem would be that on not so noisy scenes, it would smooth out way too much details.

When using QTGMC(InputType=1) is the preset="whatever" relevant? Would for example QTGMC(InputType=1, preset="fast") speed up some?