Trying to figure out if I can get rid of these "wire mesh" artifacts.

[LaserBones]

I've included an image and an unfiltered clip from the DVD I'm encoding. The DVD is 25fps PAL. Most of the footage is progressive but it does have some interlaced live action section so I'm currently using the script below. You'll notice the footage has a "wire mesh" artifact. The entire series has these artifacts. IS there anyway to resolve this?

q = QTGMC(TR0=2, TR1=2, TR2=1, Rep0=1, Rep1=0, Rep2=4, DCT=5, ThSCD1=300, ThSCD2=110, SourceMatch=3, Lossless=2, Sharpness=0.3, Sbb=0, MatchPreset="slow", NoiseProcess=2, GrainRestore=0.0, NoiseRestore=0.4, NoisePreset="slow", StabilizeNoise=false, NoiseTR=0, NoiseDeint="bob", EdiThreads=4)#.Subtitle("QTGMC") # for debugging purposes, remove subtitle for production
d = ChangeFPS(50)
ConditionalFilter(d, q, d, "IsCombedTIVTC(cthresh=2)")

[Selur]

you could for example use KNLMeansCL

KNLMeans(d=0,h=10)

and later add some line darkening, luma sharpening,...
Downside is loosing fine details, see clouds,...

Another approach would be using RemoveDirtMV and some Sharpener.


Cu Selur

Ps.: Instead of QTGMC, I would just use TFM for the combed parts.
PPs.: as alternative to KNLMeansCL one could also use HQDN3D with high values (temporal luma should be 0 or at least small).
PPs.: DPIR denoise with a strength of 25 or so will also work. (but also loose lots of small details)

[LaserBones]

you could for example use KNLMeansCL

KNLMeans(d=0,h=10)

and later add some line darkening, luma sharpening,...
Downside is loosing fine details, see clouds,...

Another approach would be using RemoveDirtMV and some Sharpener.


Cu Selur

Ps.: Instead of QTGMC, I would just use TFM for the combed parts.
PPs.: as alternative to KNLMeansCL one could also use HQDN3D with high values (temporal luma should be 0 or at least small).
PPs.: DPIR denoise with a strength of 25 or so will also work. (but also loose lots of small details)

Thank you so much! Why would you use TFM instead of QTGMC? The interlaced parts are bobbed not phase shifted. Does that matter? Would you just use

q = TFM()

in my script above?

[Selur]

Would you just use

q = TFM()

in my script above?

yes.

Why would you use TFM instead of QTGMC?

Because QTGMC seems overkill, I wouldn't want to add duplicate frames,...

[LaserBones]

Would you just use

q = TFM()

in my script above?

yes.

Why would you use TFM instead of QTGMC?

Because QTGMC seems overkill, I wouldn't want to add duplicate frames,...

Thank you!

[Selur]

If you share a sample that contains the progressive and interlaced content, maybe other settings might be better,...

[LaserBones]

If you share a sample that contains the progressive and interlaced content, maybe other settings might be better,...

Thanks Selur. I've attached a better sample. KNLMeansCL works well, but it doesn't seem to be doing a good job at the beginning of this clip. There are also 2 separate live-action shots included. Thanks!

[Selur]

Argh, the blending is in the western scene is already in the fields. (see attache images)

With the animations it seems like the chroma of some fields is shifted which TFM+KNLMeans do seem to remove rather well.
I guess during the creation of the clip they messed up some pull up/down and de-/reinerlacing.
Alternatively to TFM: QTGMC+KNLMeans+sRestore could be used, but that too will not remove those blends.

Maybe someone else has another suggestion.
I suspect that by looking at the fields one might be able to spot some patterns and fix this somewhat by undoing the patterns, but I don't see a way to get rid of the blends in the western scene.

Cu Selur

[LaserBones]

Thanks! A majority of the episodes don't have live-action scenes. So removing that residual chroma field blend is the biggest problem. I'll keep tinkering and check back here tomorrow to see if anyone else has suggestions. Thank you!