Posts: 1,539
Threads: 60
Joined: 2015 Jan
Thanks: 229
Given 632 thank(s) in 372 post(s)
Country:
Posts: 38
Threads: 6
Joined: 2015 Oct
Thanks: 1
Given 11 thank(s) in 3 post(s)
Country:
Posts: 81
Threads: 8
Joined: 2020 Oct
Thanks: 4
Given 22 thank(s) in 11 post(s)
Country:
Posts: 1,539
Threads: 60
Joined: 2015 Jan
Thanks: 229
Given 632 thank(s) in 372 post(s)
Country:
2020-11-27, 01:02 AM
(This post was last modified: 2020-11-27, 01:07 AM by Chewtobacca.)
Posts: 38
Threads: 6
Joined: 2015 Oct
Thanks: 1
Given 11 thank(s) in 3 post(s)
Country:
(2020-11-27, 01:02 AM)Chewtobacca Wrote: (2020-11-26, 11:10 PM)Kyaneos Wrote: The opening is also interlaced:
What do you suggest, check all the episodes frame by frame? 
Actually, if you separate fields, the opening isn't too bad. Each field in a given pair is marginally different from the other, but they are obviously meant to be from the same instant in time, even if you focus purely on the effects, e.g. the flight of the bird. Aliasing is possibly more of a problem. If all the effects had been done like this, the series could be made progressive without much hassle. But at the time of production, it's doubtful than anyone involved thought much about that.
Going frame by frame is probably not necessary: only certain shots have video effects work. What specifically do you want to do with the series? Is it just denoising? Depending on what you have in mind, you might get away with separating the fields, applying the filters (Dfttest, MDegrain2, QTGMC, whatever you want to use), and then weaving them back together. That would make your life easier.
Code: Mpeg2Source()
AssumeTFF()
SeparateFields()
WhateverDenoiser()
Weave()
My idea is to make the best possible encode, knowing the limitations of the material, and add the Spanish dubbing.
If it were for you, what script would you use?
I think: first to fix the interlacing, second to improve compression defects (macroblocks, banding ...), third to eliminate part of the noise, fourth to improve sharpness.
For now this is the script I'm working:
Code: QTGMC ()
SelectEven ()
Deblock_QED ()
neo_f3kdb ()
MCDegrainSharp (csharp = 0.3)
Grainfactory3 or GrainPlate or AddGrainC or f3kgrain or GradFun2DBmod or...
At the moment I don't like any grain filter. I have tried GrainPlate, with different strength levels, but the result seems artificial to me. Maybe I'm not using the best samples (35mm Digital Grain from https://grainzilla.com/free-film-grain-download/ and FILMGRAIN_HD_35MM_25FPS_PRORES_422.mov from https://drive.google.com/drive/folders/0...QTZ1ZXFlcX ). I have used the two functions that I have found here:
Code: function grainplate (clip clip, "strength") {
strength = default (strength, 0.25)
g = lsmashvideosource ("1080P-35MM-DIGITAL-GRAIN.mov", format = "YUV422P8")
g = MergeRGB (g, UnalignedSplice (g.trim (g.framecount / 2-1, g.framecount-1), g.trim (0, (g.framecount / 2) -1)), UnalignedSplice (g.trim (g.framecount / 3-1, g.framecount-1), g.trim (0, (g.framecount / 3) -1)))
grain = g.crop (0,0,720,576) .loop.trim (0, clip.framecount-1)
return overlay (clip.tweak (bright = 1), grain, mode = "softlight", opacity = strength)}
Code: function grainplate (clip clip, "strength") {
strength = default (strength, 0.25)
grain = lsmashvideosource ("FILMGRAIN_HD_35MM_25FPS_PRORES_422.mov", format = "YUV422P8") \
.crop (0,0,720,576) .loop.trim (0, clip.framecount-1)
return overlay (clip.tweak (bright = 1), grain, mode = "softlight", opacity = strength)}
Another thing I don't like is the mosquito noise amplification that MCDegrainSharp does around the credits:
My knowledge of Avisynth is limited, I accept suggestions about filters, order of filters, options etc.
Posts: 1,539
Threads: 60
Joined: 2015 Jan
Thanks: 229
Given 632 thank(s) in 372 post(s)
Country:
Posts: 38
Threads: 6
Joined: 2015 Oct
Thanks: 1
Given 11 thank(s) in 3 post(s)
Country:
|