NTSC DVD w/ hard lines and duplicate frames - deinterlace/detelecine?

[U2Joshua]

Sample: https://www.sendspace.com/file/ja2zzx

This is an NTSC DVD that appears to have every 4th frame duplicated, as well it appears to have interlacing lines (to my eye), but DGIndex calls the video progressive.

Using MeGUI + AVISynth I'm trying to figure out what can be done. I'm assuming decimate might be able to rid the duplicate frames, but is there away to remove the apparent interlacing lines?

Is this as simple as QTGMC + decimate? Or, is this hard interlacing and may either require special treatment, or, perhaps nothing can be done about it at all?

Thanks in advance~

[hello_hello]

Someone may have a better idea and I don't really know how it got the way it is, but QTGMC seems to clean it up reasonably well. I don't think it's actually interlaced though. Something like this maybe....

TDecimate()
QTGMC()
SelectEven()
crop(20, 14, -20, -14)
Spline36Resize(832,468,1,0,-1,0)
LSFMod()

[JVRaines]

It's not interlaced. Looks like 24 fps source with duplicated frames to make it 29.97 and some blends crept in along the way.

[manono]

Yeah, I'd do something similar to hello_hello, maybe QTGMC in progressive mode after removing the dupes. It doesn't look like real interlacing to me, either, but maybe aliasing.

[U2Joshua]

Yeah, I'd do something similar to hello_hello, maybe QTGMC in progressive mode after removing the dupes. It doesn't look like real interlacing to me, either, but maybe aliasing.

Thanks, everyone... what's the code for QTGMC progressive?

[hello_hello]

QTGMC(InputType=1)
QTGMC(InputType=2)
QTGMC(InputType=3)

All the above are progressive modes. I can't remember the exact differences. You'd need to check the help file. It doesn't work for your video though. I tried InputType=1 and it made the aliasing worse. One of the other progressive modes might do slightly better but I'm pretty sure de-interlacing followed by SelectEven() to remove the repeated frames will give you the best result.

Edit: It was annoying me, not being able to remember, so I looked it up.

This script is designed for deinterlacing and so by default expects an interlaced clip. However, much of its operation concerns the reduction of horizontal shimmering. It is possible to use the script to remove similar shimmer from a progressive clip by using the InputType setting. InputType=1 is used for general progressive material that contains less severe problems. InputType=2,3 are designed for badly deinterlaced material with considerable horizontal artefacts. These modes drop half the fields and recreate an interlaced stream from the others, which is then treated in the normal way. Mode 3 differs from mode 2 only in that it complements field parity of the input. Generally mode 1 will retain more detail, but repair less artefacts than modes 2,3. You may consider setting TR2 to a higher value (e.g. 2 or 3) when repairing progressive material.

[JVRaines]

Gory details.
http://avisynth.nl/index.php/QTGMC#Progressive_Input

[U2Joshua]

Gory details.
http://avisynth.nl/index.php/QTGMC#Progressive_Input

Thanks

...maybe one of these days I'll learn how to read for myself haha...

[jagabo]

The video isn't interlaced at all. It's just been poorly deinterlaced. And from the blending artifacts it was probably originally PAL video.

[U2Joshua]

Mode 3 differs from mode 2 only in that it complements field parity of the input.

Can anyone tell me what that means?

...You may consider setting TR2 to a higher value (e.g. 2 or 3) when repairing progressive material.

...as well as this?

Thanks in advance~

I've been messing around with several of these suggested settings and am seeing some great results!

[hello_hello]

Mode 3 differs from mode 2 only in that it complements field parity of the input.

Can anyone tell me what that means?

The way I read it, mode 2 removes every second scanline in a frame (described as "dropping half the fields"), they're replaced by interpolating new scanlines from the remaining ones, then the video is de-interlaced. I assume if mode 2 keeps all the even numbered scanlines, mode 3 must keep the odd ones.

...You may consider setting TR2 to a higher value (e.g. 2 or 3) when repairing progressive material.

...as well as this?

TR2 (0,1,2,3) Temporal linear smoothing radius used for final stablization / denoising. Increase for smoother output

The defaults for many settings change if you change the speed preset, and the default speed preset is "slower" which sets TR2=1.

You can see the currently used settings by previewing a script like this:

QTGMC(InputType=1, ShowSettings=true)

If you were to specify the fast preset you'd see it changes to TR2=0

QTGMC(InputType=1, Preset="fast", ShowSettings=true)

You can over-ride individual settings independently of any speed preset.

QTGMC(InputType=1, Preset="fast", TR2=3, ShowSettings=true)

The presets are a preconfigured trade off between speed and quality (default "slower").
Preset (string) Sets a range of defaults for different encoding speeds. Select from "Placebo", "Very Slow", "Slower", "Slow", "Medium", "Fast", "Faster", "Very Fast", "Super Fast", "Ultra Fast" & "Draft"

If you want extra denoising, you use the EZDenoise function with a value greater than zero. The value sets the denoising strength.

QTGMC(InputType=1, Preset="medium", ShowSettings=true, EzDenoise=1.5)

The EzDenoise function has it's own speed presets somewhat independent of the main speed presets. The default is "fast" which is pretty much all I ever use. There's no need to specify either speed preset unless you want to use something other than the defaults.

QTGMC(InputType=1, Preset="medium", ShowSettings=true, EzDenoise=1.5, NoisePreset="medium")

NoisePreset (string) Automatic setting for quality of noise processing. Choices: "Slower", "Slow", "Medium", "Fast", and "Faster", Default is "Fast"

For most presets the defaults are: NoiseProcess=0, Sigma=2.0, NoiseRestore/GrainRestore=0.0, ChromaNoise=false, DftThreads=EdiThreads, ShowNoise=false
In "Placebo", "Very Slow" some defaults change: NoiseProcess=2, GrainRestore=0.3, NoiseRestore=0.1

EZDenoise and EZKeepGrain set automatic values for NoiseProcess, NoiseRestore/GrainRestore and Sigma
Denoiser, DenoiseMC, NoiseTR, NoiseDeint and StabilizeNoise settings are determined by NoisePreset
If using manual settings be sure to override *all* the values you are interested in

[jagabo]

Something like this will reduce the aliasing better and will be much faster than QTGMC:

Mpeg2Source("War-No More Trouble.d2v", CPU2="ooooxx", Info=3) 
TDecimate()
Deblock_qed() # fine tune this
Spline36Resize(width,288) # back to PAL field height
nnedi3(dh=true) # double the height with smooth edges
Spline36Resize(width,480) # back to NTSC frame height
Sharpen(0.0, 0.6) # sharpen vertically

You could always use QTGMC after this.

[U2Joshua]

Something like this will reduce the aliasing better and will be much faster than QTGMC:

Mpeg2Source("War-No More Trouble.d2v", CPU2="ooooxx", Info=3) 
TDecimate()
Deblock_qed() # fine tune this
Spline36Resize(width,288) # back to PAL field height
nnedi3(dh=true) # double the height with smooth edges
Spline36Resize(width,480) # back to NTSC frame height
Sharpen(0.0, 0.6) # sharpen vertically

You could always use QTGMC after this.

Thanks, @jagabo. By "you could always use QTGMC after this." do you mean that I would run your script, and run a new script on the processed video based on your script? Or, simply add in QTGMC after your last line? Either way, would I still want to / need to indicate Selecteven in order to keep fps in check?

[jagabo]

By "you could always use QTGMC after this." do you mean that I would run your script, and run a new script on the processed video based on your script? Or, simply add in QTGMC after your last line?

Just add QTGMC at the end of that script. You might want to reduce the Sharpen() because QTGMC will sharpen a bit.

Either way, would I still want to / need to indicate Selecteven in order to keep fps in check?

No, use QTGMC(InputType=1) (and whatever other options you might want) so you get just the cleanup, not the bobbing.