Unable to remove all of the interlacing on Blu-Ray - Anime

[AOA]

Hello. In the past I've filtered a handful of dvds and some blurays, but I've never come across interlacing this stubborn:

raw file @ mega.nz (https://mega.nz/file/VYMmSBpJ#-E7P8kHsInKu_t-aWEWAV7YBL34nMfvXjyC-_sRD2xM)

Most of the interlacing can be removed without any problems, but at time "00:00:32.299" f.e. it's still left.
I've realised that those artficats look bigger/different, so maybe it's the fault of the mastering studio?
Can it still be fixed?


IVTC part of my script:

clip = core.dgdecodenv.DGSource(source)
def postprocess(n, f, clip, deinterlaced):
if f.props['_Combed'] > 0:
return deinterlaced
else:
return clip
matched_clip = core.vivtc.VFM(clip, order=1, field=0, cthresh=4, mode=4, micout=True)
deinterlaced_clip = core.eedi3m.EEDI3(matched_clip, field=0)
postprocessed_clip = core.std.FrameEval(matched_clip, functools.partial(postprocess, clip=matched_clip, deinterlaced=deinterlaced_clip), prop_src=matched_clip)
decimated_clip = core.vivtc.VDecimate(postprocessed_clip)

Yep, some of the VFM options are unsafe; I just tried everything to fix this.
(including QTGMC instead of EEDI3)

I would really appreciate if someone could give me a hint :)

[Myrsloik]

Simple answer: the crystal section in your example is just broken. Probably due to odd framerate conversions done during editing of the show.

If you separate the fields you see that every field is repeated (not telecined) so then decimating half the frames and encoding it as 15 fps seems like a good idea. (if you don't mind the complexity of VFR encoding)

There are however two more problems. Even if you separate the fields they individually appear to have combing/interlacing artifacts in them as well. It's an extremely mangled source.
And if you put both top and bottom fields together they're not from the same point in time which is pure nonsense if the fields themselves are repeated. This can't have looked smooth on a crt tv in the old days either.

What I'd try:
Figure out which field is the better one (different number of combing artifacts, chroma ghosting and such).
Drop every other frame.
Re-interpolate the dropped field with eedi3/nnedi3/whatever
Encode this section as 15 FPS VFR.

Maybe it'll look less bad that way. It will never look good.

[AOA]

Simple answer: the crystal section in your example is just broken. Probably due to odd framerate conversions done during editing of the show.

If you separate the fields you see that every field is repeated (not telecined) so then decimating half the frames and encoding it as 15 fps seems like a good idea. (if you don't mind the complexity of VFR encoding)

There are however two more problems. Even if you separate the fields they individually appear to have combing/interlacing artifacts in them as well. It's an extremely mangled source.
And if you put both top and bottom fields together they're not from the same point in time which is pure nonsense if the fields themselves are repeated. This can't have looked smooth on a crt tv in the old days either.

What I'd try:
Figure out which field is the better one (different number of combing artifacts, chroma ghosting and such).
Drop every other frame.
Re-interpolate the dropped field with eedi3/nnedi3/whatever
Encode this section as 15 FPS VFR.

Maybe it'll look less bad that way. It will never look good.

Thanks for the throughough explaination.
Makes a lot more sense.
Unfortunately, this isn't the only thing the studio messed up while mastering this.
Other frames have only the chroma layer interlaced in a similar way, which I only found out after posting this.
If anyone wants to waste their time on another qtec, its name is "Ai no Kusabi (2012)".
Cool another dropped project because of incompetent humans :(