Reasonable filtering of VHS capture?

[sventamyra]

Hi

Thanx for a great forum!

I'm setting up to capturing the old family VHS films. And chose the wifves work out video as a starter.
I bought a used JVC HM-DR10000EU (with line TBC) connected via s-video to a conexant polaris video USB-dongle in a win7x64 laptop. I'm captureing i Virtualdub with lagarith YUY2.

I was thinking of running the follwing avisynth-script to make it watchable on my 65" samsung. I tried having the tv doing the de interlacing in runtime but found deinterlacing looking better.

AviSource("VHS1.avi")  
ConvertToYV12(interlaced=true)  
AssumeTFF()
QTGMC(Preset="very Slow", EzDenoise=0.0, EdiThreads=1) #Preservefields=true?
Crop(12, 2, -16, -12) 
Levels(18, 1.02, 235, 16, 235, dither=true, coring=false) #try to use the complete 16->235 range
ColorYUV(off_u=+3, off_v=-3) #Checked some histograms and tried to adjust the colors
Tweak(hue=-3, sat=0.85, coring=false, dither=true)

#DeScratch()  
#Turnright().RemoveDirtMC_SE(GPU=false, radius=5, repmode=0, clmode=0, TwoPass=false, thSAD=35, thSADC=35).turnleft()

LSFmod(strength=130)  
#Stab(range=1, dxmax=1, dymax=1) #Thought I saw the video bump one line somewhere. Cant find it anymore
#prefetch(1)
AddBorders(12, 8, 10, 12)  
return last

1. There are some screatches/comets. Never really got to get rid of them without making it look too filtered. Any good ideas? Theres two in the samples

2. What level of luma noice /chroma noise filtering would you apply?
Would you filter for spots? There some.

3. I was thinking of getting an external frame TBC (panasonic DVR). Would it enhance this video?

4. Sample 23067 contains a weired aspect distorsion. The film is full of intentional effects to mek it look arty. Would you say that this is intentional? Or is it my VCR/tape that's bugging me?

Cheers!

[Aya_Rei]

About the warp sample, I'm guessing it's part of the film because it doesn't affected the added in subtitles, head switching noise and that the frame itself looks stable

[sventamyra]

Good point. Thanx

[sventamyra]

What about question 1? I ca t seem to get rid om those transparent tears. Usually they are still for 1-2 frames. But this one moves. Do I need to replace the conplete frame?

[keaton]

Attached test script just to remove the comets. It uses RemoveDirtMC.avs, which I attached in case you don't have that. You may need to install more plugins to make RemoveDirtMC work. Also attached the resulting file in HuffYUV format.

[traal]

I didn't notice any scratches/comets.

Generally, I avoid NR unless the noise is distracting, which it isn't in the three videos posted. The reason I avoid it is because it makes people look fake, which is worse than a little noise.

The warping in 23067 looks like a line-TBC issue, but the left edge of the video appears to be masked out so it's hard to tell for sure.

[sventamyra]

Quote:

Originally Posted by keaton

Attached test script just to remove the comets. It uses RemoveDirtMC.avs, which I attached in case you don't have that. You may need to install more plugins to make RemoveDirtMC work. Also attached the resulting file in HuffYUV format.

Thank you so much!. I was running RemoveDirtMC after QTGMC and it wasn't working at all. You Clean-function cleared that for me.

My results was overall looking quite plastic. So I've been trying to cutback on filtering. Found someone here that played around with QTGMC and recommended:

QTGMC(TR0=2, TR1=2, TR2=1, Rep0=1, Rep1=0, Rep2=4, DCT=5, ThSCD1=300, ThSCD2=110, SourceMatch=3, Lossless=2, Sharpness=0.1, Sbb=0, MatchPreset="slow", NoiseProcess=2, GrainRestore=0.0, NoiseRestore=0.4, NoisePreset="slow", StabilizeNoise=false, NoiseTR=0, NoiseDeint="bob", Border=true)

I think it made a little difference.
Also tried w/o deinterlacing, but it looked terrible on my TV.
And then I decided that removedirtMC also added too much filtering, så I created a mask from its detections and included a threashold:

diff=mt_makediff(original, fixedclip, U=0, V=0).convertToY8().ConvertToYV12(interlaced=false)
diff=mt_lut(diff, expr="x 128 - abs").mt_expand(U=3, V=3).mt_expand(U=3, V=3).mt_expand(U=3, V=3)#.mt_expand(U=3, V=3)
mask=mt_lut(diff, expr="x 32 > 255 0 ?")# Binär mask
scale_h=8
scale_w=60
mask=mask.bicubicresize(mask.width/scale_w, mask.height/scale_h)
mask=mask.bicubicresize(mask.width*scale_w, mask.height*scale_h)
mask=mt_lut(mask, expr="x 14 *")
fixedclip = mt_Merge(original, fixedclip, mask)

I think it helped a little.The mask looks like this: The grey is what RemovedirtMC changed, and the black is the threasholded mask created