Deinterlacing issue and green line

[baka1]

I have a deinterlacing issue with this ancient PAL DVD. Avisynth's Analyse suggests using Yadifmod2(order=1), however, I get this random green line that pops up 2 pixels on the bottom half of the screen.

https://i.ibb.co/gMRx5HWt/Combined024242.png (https://ibb.co/jvrScBw3)

Here are the sample clips with two screenshots, one with the green line issue above, another with the ghosting image using QTGMC and TFM.

https://mega.nz/folder/XKxFkTLL#bD6vfs2SqdeJWc2Bx4vTGA

[DJ Bobo]

Could it be that you are cropping before deinterlacing?
Always deinterlace first.
Cropping should be preferably in multiples of 4 btw.

[baka1]

Oooh yes, moving cropping at the end of the script removes the green line issue.

Still can't find the proper filter to deinterlace it, tho.

[Selur]

quick look: QTGMC+sRestore(23.976) seems to work fine

[baka1]

Blended frames still show: (with QTGMC, no srestore)
https://i.ibb.co/35rGc7sv/New-File-1-000260.png (https://ibb.co/nqgZjb0B)
https://i.ibb.co/mFVdbQhs/New-File-1-000129.png (https://ibb.co/n8N46hQ2)

With srestore 23.976, I find duplicated frames (266=267; 272=273; 279=280), which don't appear at 25fps.

The sequence from 260 to 281, seems to follow a telepathic ghost frame, showing a glimpse of future movement, followed by a frame of actual movement, etc etc.

[DJ Bobo]

@ baka1
This is an interpolated PAL DVD, so very hard to deinterlace properly without leaving any kind of ghosting or blur, as the blending happens at field level.

[Selur]

Sadly, I don't see a way to do this automatically.
Duplicate frames could be replaced with interpolations, but I don't see a way to fix the blending in the fields.

Cu Selur

[baka1]

Thank you DJ Bobo and Selur (regular user of Hybrid).

As I feared, the interpolated PAL DVD has unfixable blended fields.

[Sharc]

With srestore 23.976, I find duplicated frames (266=267; 272=273; 279=280), which don't appear at 25fps.
Sidenote: It's the other way round here: duplicates with QTGMC().SRestore(frate=25.0) and no duplicates for QTGMC().SRestore(frate=23.976). But unfortunately for either case only some of the blends get fixed by SRestore for reasons mentioned before.