Dealing with old DVD with very bad Image

[ravarage]

one of the anime that I love is NANA (2006), I have R2J DVD (NTSC,30000/10001,868x480), those DVD suffer a lot of banding, dot noise, I used vapoursynth to fix interlace,and balance between banding and dithering

any suggestion

 import vapoursynth as vs
import sys
import importlib
import functools
core = vs.get_core()
sys.path.append(r"C:\vs\VapourSynth64Portable\Scripts")
ASTDR = importlib.machinery.SourceFileLoader("ASTDR", r"C:\vs\VapourSynth64Portable\Scripts\ASTDR.py").load_module()
havsfunc = importlib.machinery.SourceFileLoader("havsfunc", r"C:\vs\VapourSynth64Portable\Scripts\havsfunc.py").load_module()
G41Fun = importlib.machinery.SourceFileLoader("havsfunc", r"C:\vs\VapourSynth64Portable\Scripts\G41Fun.py").load_module()
NRDB = importlib.machinery.SourceFileLoader("havsfunc", r"C:\vs\VapourSynth64Portable\Scripts\NRDB.py").load_module()
import mvsfunc as mvf
clip = core.ffms2.Source(r"F:\0\15short/13-002.mkv",fpsnum=30000, fpsden=1001)
clip = havsfunc.QTGMC(clip, Preset='Very Slow', SourceMatch=3, TR2=1,Rep2=1, Lossless=2, TFF=True,FPSDivisor=2,NoiseProcess=1, NoiseRestore=0.0, Denoiser="dfttest", DenoiseMC=True, NoiseTR=2, Sigma=4.0  )
clip = core.placebo.Resample(clip,width=864,height=480, filter = "ewa_lanczos", sx = 0.0, sy = 0.0, antiring = 0.0, lut_entries = 64, cutoff = 0.001, sigmoidize = 1, linearize = 1, sigmoid_center = 0.75, sigmoid_slope = 6.5, trc = 1)
clip = G41Fun.SuperToon(clip)
clip = core.dfttest.DFTTest(clip ,  ftype=0,  sigma=8.0,  sigma2=8.0,  pmin=0.0,  pmax=500.0,  sbsize=16,  smode=1,  sosize=12,  tbsize=3,  tmode=0,  tosize=0,  swin=0,  twin=7,  sbeta=2.5,  tbeta=2.5,  zmean=True,  f0beta=1.0,  nlocation=None,  slocation=None,  ssx=None,  ssy=None,  sst=None,  ssystem=0,  planes=[0, 1, 2],  opt=0)
clip = NRDB.NRDB(clip, deband="", db_Y=64, db_Cb=0, db_Cr=0, grainY=0, grainC=0, dynamic_grain=False, keep_tv_range=False, nr="", planes=[0], sigma=4.0, sigma2=4.0)
super =  core.mv.Super(clip,pel=2, rfilter =4 )
bv1 = core.mv.Analyse(super, isb = True, delta = 1, overlap=4)
fv1 = core.mv.Analyse(super, isb = False, delta = 1, overlap=4)
bv2 = core.mv.Analyse(super, isb = True, delta = 2, overlap=4)
fv2 = core.mv.Analyse(super, isb = False, delta = 2, overlap=4)
bv3 = core.mv.Analyse(super, isb = True, delta = 3, overlap=4)
fv3 = core.mv.Analyse(super, isb = False, delta = 3, overlap=4)
clip =  core.mv.Degrain3(clip,super,bv1,fv1,bv2,fv2,bv3,fv3,thsad=400)
clip = havsfunc.HQDeringmod(clip, nrmode=2, darkthr=3.0)
clip = havsfunc.FastLineDarkenMOD(clip, strength=48, protection=5, luma_cap=191, threshold=4, thinning=0)
clip.set_output()

[KarMa]

Post a short sample directly to this site.

[ravarage]

here is source and sample encode

[manono]

any suggestion

You should IVTC and not deinterlace.

TFM()
TDecimate(Mode=1)

[ravarage]

any suggestion

You should IVTC and not deinterlace.

TFM()
TDecimate(Mode=1)

I used that but the frame jump back and forward too much, what advantages of this over the QTMC that I used ?

[davexnet]

Tdecimate provides the underlying 24 fps, rather than the 30 with duplicate frames.
Use it instead of the qtgmc.

[manono]

I used that but the frame jump back and forward too much...

You had the field order wrong. Use AssumeTFF (or AssumeBFF) before the TFM call to correct it.

what advantages of this over the QTMC that I used ?

It will restore the framerate to its original and correct 23.976fps.

[davexnet]

I used that but the frame jump back and forward too much...

You had the field order wrong. Use AssumeTFF (or AssumeBFF) before the TFM call to correct it.

what advantages of this over the QTMC that I used ?

It will restore the framerate to its original and correct 23.976fps.

Tried MCTemporalDenoise (default settings) perhaps this will help.
Although at the 14-15 seconds mark, the OP's script seemed to smooth the posterization
a little better

[jagabo]

If you just want to reduce the posterization you can use GradFun3() or GradFun2DBMod(). Then you have to encode with enough bitrate to retain the added noise. Otherwise you might as well not bother.

[ravarage]

much better any other suggestions ?

import vapoursynth as vs
import sys
import importlib
import functools
core = vs.get_core()
sys.path.append(r"C:\vs\VapourSynth64Portable\Scripts")
havsfunc = importlib.machinery.SourceFileLoader("havsfunc", r"C:\vs\VapourSynth64Portable\Scripts\havsfunc.py").load_module()
mvs = importlib.machinery.SourceFileLoader("havsfunc", r"C:\vs\VapourSynth64Portable\Scripts\mvsfunc.py").load_module()
G41Fun = importlib.machinery.SourceFileLoader("havsfunc", r"C:\vs\VapourSynth64Portable\Scripts\G41Fun.py").load_module()
NRDB = importlib.machinery.SourceFileLoader("havsfunc", r"C:\vs\VapourSynth64Portable\Scripts\NRDB.py").load_module()
clip = core.ffms2.Source(r"F:\testarea/04-006.mkv",fpsnum=30000, fpsden=1001)
clip = core.tcomb.TComb(clip, mode=2)
clip = mvs.AssumeTFF(clip ) 
matched_clip = core.vivtc.VFM(clip,1,cthresh=10)
clip= core.vivtc.VDecimate(matched_clip)
clip = core.placebo.Resample(clip,width=864,height=480, filter = "ewa_lanczos", sx = 0.0, sy = 0.0, antiring = 0.0, lut_entries = 64, cutoff = 0.001, sigmoidize = 1, linearize = 1, sigmoid_center = 0.75, sigmoid_slope = 6.5, trc = 1)
clip = havsfunc.MCTemporalDenoise(clip)
clip = core.f3kdb.Deband(clip)
super =  core.mv.Super(clip,pel=2, rfilter =4 )
bv1 = core.mv.Analyse(super, isb = True, delta = 1, overlap=4)
fv1 = core.mv.Analyse(super, isb = False, delta = 1, overlap=4)
bv2 = core.mv.Analyse(super, isb = True, delta = 2, overlap=4)
fv2 = core.mv.Analyse(super, isb = False, delta = 2, overlap=4)
bv3 = core.mv.Analyse(super, isb = True, delta = 3, overlap=4)
fv3 = core.mv.Analyse(super, isb = False, delta = 3, overlap=4)
clip =  core.mv.Degrain3(clip,super,bv1,fv1,bv2,fv2,bv3,fv3,thsad=400)
clip = havsfunc.HQDeringmod(clip, nrmode=2, darkthr=3.0)
clip = havsfunc.FastLineDarkenMOD(clip, strength=48, protection=5, luma_cap=191, threshold=4, thinning=0)
clip = G41Fun.SuperToon(clip)
clip.set_output()