QTGMC Deinterlacing Script (v3.384)

[TDS]

Originally Posted by flossy_cake

[deleted - redundant]

Hey flossy, if you want to actually delete a message simply go into Edit then in the bottom right corner there's Delete option then you go over to the bottom left corner Delete this message, and then go back over to the right corner and delete it.

[Citaln]

Trying to post again, since I'm not sure if my post got lost in the moderation queue. I have some edge artifacts with QTMC that I tried to solve with the border option and I just don't understand what border does or why; it seems like the top/bottom rows are 50% of the previous field. Is that intentional? For example:

White = BlankClip(length=1,width=40,height=160,color=$FFFFFF)
Black = BlankClip(length=1,width=40,height=160,color=$000000)

BlankClip(length=1,width=40,height=160,color=$FFFFFF)
StackHorizontal(Last,Black)
StackHorizontal(Last,Black)
Frame1 = StackHorizontal(Last,Black)

BlankClip(length=1,width=40,height=160,color=$000000)
StackHorizontal(Last,White)
StackHorizontal(Last,Black)
Frame2 = StackHorizontal(Last,Black)

BlankClip(length=1,width=40,height=160,color=$000000)
StackHorizontal(Last,Black)
StackHorizontal(Last,White)
Frame3 = StackHorizontal(Last,Black)

BlankClip(length=1,width=40,height=160,color=$000000)
StackHorizontal(Last,Black)
StackHorizontal(Last,Black)
Frame4 = StackHorizontal(Last,White)

Interleave(Frame1,Frame2,Frame3,Frame4)

SeparateFields().SelectEvery(4,0,3).Weave()

converttoYV16

qtgmc(border=true)

With border=false, deinterlace is done correctly, with border=true I get problem mentioned, which is on the horizontal axis so why would a vertical padding do that?

[flossy_cake]

Originally Posted by TDS

Hey flossy, if you want to actually delete a message simply go into Edit then in the bottom right corner there's Delete option then you go over to the bottom left corner Delete this message, and then go back over to the right corner and delete it.

Thanks - done.

On a separate note I was wondering if anyone familiar with the QTGMC source code could say definitively whether the EdiThreads parameter is used when InputType==1?

Looking at QTGMC.avsi it seems like it just returns the input clip without doing any EDI and thus not using the EdiThreads param:

function QTGMC_Interpolate( clip Input, int InputType, string EdiMode, int NNSize, int NNeurons, int EdiQual, int EdiMaxD, int EdiThreads, clip "Fallback", string "ChromaEdi" )
{
	interp = (InputType == 1)            ? Input :

In this case I'm curious to know what InputType==1 is actually doing? I thought it still needed to blend or compare the input clip with EDI clip to know where to smooth edges?

Originally Posted by Citaln

With border true, the bottom and top lines are just plain 50% of the previous or next field. Is this correct/intended?

I think so, because QTGMC uses temporal filtering that blends with previous/next frame as part of its antialiasing/deshimmering magic. So I think some of that blending is bleeding through into the previous/next frame at the top and bottom where it was blanked off.

Perhaps what you could do instead is blank off the top and bottom pixel rows with Letterbox(1, 1, 0, 0) and then simply copy paste the neighbouring row into it. Although I am not sure how to do that 2nd step - I usually just zoom the image in by that many pixels, but that may not be desirable if you want to preserve the pixel alignment...

/*------------------ Vertical zoom ------------------*/

zoomY = 2.0 # pixels to zoom in vertically

Spline36Resize(width, height, 
\  zoomY*(float(width)/float(height)),    /*left*/
\  zoomY,                                 /*top*/
\ -zoomY*(float(width)/float(height)),    /*right*/
\ -zoomY)                                 /*bottom*/


/*----------------- Horizontal zoom -----------------*/

zoomX = 2.0 # pixels to zoom in horizontally
	 	
Spline36Resize(width, height, 
\  zoomX,                                 /*left*/
\  zoomX*(float(height)/float(width)),    /*top*/
\ -zoomX,                                 /*right*/
\ -zoomX*(float(height)/float(width)))    /*bottom*/

[BilboFett]

new Avisynth+ 3.7.3. broke my QTGMC.
Everything is in their proper folders and paths, I double checked.
Even after uninstalling, rebooting, reinstalling 3.7.2, rebooting, ..I still get this message.



I even tried importing the QTGMC.AVSI file
Everything worked fine before the upgrade.
It still works fine inside Vapoursynth imported into Avisynth.
Nothing else has changed on this system.
Please help, TIA
(cross-posting to Avisynth thread)

[FranceBB]

@BilboFett... it's probably something related to your environment or your build.

[kedautinh12]

Or conflict with 2 or more function name QTGMC in your same folder plugins. Check with avsmeter

[DTL]

If importing file also not work - you can copy and paste QTGMC function to the beginning of the script manually.

[LigH]

Besides, DirectShowSource is not the source filter I would recommend primarily for MKV files...

But yes, first test both the general environment and the processing of this script using AVSMeter.

[BilboFett]

Originally Posted by DTL

If importing file also not work - you can copy and paste QTGMC function to the beginning of the script manually.

do you see the screenshot? I'm not sure what you're telling me to do.. is the function of QTGMC not there in my script, near the beginning? I can't have it as the very 1st line, of course

[BilboFett]

Originally Posted by kedautinh12

Or conflict with 2 or more function name QTGMC in your same folder plugins. Check with avsmeter

nope, only one function named QTGMC inside only one folder. Same as before when it was working prior to the upgrade.

[BilboFett]

Originally Posted by LigH

Besides, DirectShowSource is not the source filter I would recommend primarily for MKV files...

But yes, first test both the general environment and the processing of this script using AVSMeter.

What would you recommend for source filter? FFVideoSource?
I'll use Avsmeter and report back

[kedautinh12]

He said copy all of text from qtgmc.avsi to beginning of the script. After use your scripts below

[kedautinh12]

Originally Posted by BilboFett

nope, only one function named QTGMC inside only one folder. Same as before when it was working prior to the upgrade.

You need check with Avsmeter for sure. If two or more duplicate functions name, Avsmeter will announced you

[kedautinh12]

Originally Posted by BilboFett

What would you recommend for source filter? FFVideoSource?
I'll use Avsmeter and report back

Ffms2, L-SMASH-Works or DGindexNV will better

[BilboFett]

Originally Posted by kedautinh12

You need check with Avsmeter for sure. If two or more duplicate functions name, Avsmeter will announced you

There is only one function named, only one QTGMC in one single folder, just as before the upgrade when it worked fine.

[real.finder]

Originally Posted by BilboFett

There is only one function named, only one QTGMC in one single folder, just as before the upgrade when it worked fine.

try with simple QTGMC() (without any parameters)

[LigH]

I found a clip QTGMC has trouble with: It works nicely when a diagonal line is rather vertical (>50°), but when the camera turns so that this line becomes rather horizontal (<40°), the result is very jagged. Are there any parameters which can help avoiding that?

I tried different presets, tested with different SourceMatch and Lossless parameters (which are supposed to help against aliasing). Did not yet find a cure against a gentle slope being turned into a sawtooth line; steep slopes are straightened pretty well though.

The unprocessed material looks better in this scene than the deinterlaced result by QTGMC. Is there a threshold to keep more original material when only little motion is detected?
__

Question cancelled. Apparently the "bad scene" is progressive, between interlaced scenes.

[flossy_cake]

Originally Posted by LigH

Question cancelled. Apparently the "bad scene" is progressive, between interlaced scenes.

Try InputType=1 which is for antialiasing progressive material, check out this demo: https://drive.google.com/file/d/1D4P...ew?usp=sharing

Make sure to download the file and play it in your media player not in web browser which is Google's transcode

[LigH]

Good to know for the next time; but this material revealed several more issues, so I decided it's "beyond repair".

[BilboFett]

Originally Posted by real.finder

try with simple QTGMC() (without any parameters)

same thing; "there is no function known as QTGMC"