ghosting after IVTC & QTGMC

[clashradio]

Hello,

I'm having trouble with ghosting. I've tried a bunch of options with various scripts.

Project is a composite of NTSC laserdisc & BD, so delivery will be 1080 24p.

Pay attention to Stewart's nose as he moves his head.

Current script I've got the best result is:

v= LWlibavVideoSource("side_1.mov",format="YUY2")
a= LWlibavAudioSource("side_1.mov")
AudioDub(v,a)
AssumeTFF()
QTGMC(preset="slow")
SelectEvery(5,2,4)
prefetch(8)
ColorMatrix(mode="rec.601->rec.709")
nnedi3_rpow2(2,cshift="Spline36Resize",fwidth=1920 ,fheight=1440)

Footage is 16mm, so I originally tried this script:

v= LWlibavVideoSource("side_1.mov",format="YUY2")
a= LWlibavAudioSource("side_1.mov")
AudioDub(v,a)
TFM()
TDecimate(0)
prefetch(8)
ColorMatrix(mode="rec.601->rec.709")
nnedi3_rpow2(2,cshift="Spline36Resize",fwidth=1920 ,fheight=1440)

Any ideas/changes in the script is much appreciated.

[davexnet]

Try

qtgmc()
srestore(frate=23.976)

[clashradio]

Hi davexnet. I tried that code and it is an improvement thank you. Looks like frames 16-18, 31-33 are still hairy.

Is there a way to adjust the script and see the instant results while the .avs is open in VirtualDub? or do you always have to reload the .avs?

[davexnet]

Hi davexnet. I tried that code and it is an improvement thank you. Looks like frames 16-18, 31-33 are still hairy.

Is there a way to adjust the script and see the instant results while the .avs is open in VirtualDub? or do you always have to reload the .avs?

In Vdub2, press f2 to reload the script. The may be some other options using AvsPmod, but I'm not that famiuliar with it

[jagabo]

It's F2 if the main VirtualDub2 window has the focus. If the script window has the focus it's F5. One of the annoying things about VirtualDub2.

It's F5 in AvsPmod.

[clashradio]

Thanks davexnet & jagabo for the VirtualDub help.

Am I stuck with this blending/ghosting? Is this baked in the video due to an improper IVTC?

[jagabo]

Am I stuck with this blending/ghosting?

Probably.

Is this baked in the video due to an improper IVTC?

More likely a frame rate conversion. Or multiple frame rate conversions. Maybe a temporal noise reduction filter too.

SRestore(frate=25.0) may work a little better. A long panning shot would be better for analyzing the correct frame rate. Also note that SRestore can take a while to lock into a pattern. So if you just scrub around the video and look at random frames you will see a lot more residual blending than if you step through a video frame by frame.

[clashradio]

Thanks jagabo. Looks like SRestore(frate=25) gives the best results. Problem is I converted the BD footage to 24p. The m2ts was originally 29.97 interlaced, which is confusing because they restored the 16mm footage for the 2022 BD, assumably from the negative which should be 24fps. I would think the BD should be 23.976.