VHS Extraction has grid Pattern over it

[killerteengohan]

I'm not sure what this is, but the extraction from VHS tape has come out with a diamond shaped grid like pattern all over it.

Anything I can do in avisynth to fix it or make it less obvious?

Here is a 30 second video sample.
https://mega.nz/file/IpJX0CwC#WagWbnkfvFTUH-ty9aONgBabDi2Q7TOLBSLu6FC1Wts

[dellsam34]

Was that noise available in the tape or caused by bad connection during capturing? I've seen bad TV reception could cause that so most likely recorded that way, There maybe some restoration filters that can eliminate such noise but will affect the sharpness of the video.

[johnmeyer]

I have the same question about whether it was on the tape prior to the transfer.

It looks like the cross-hatch pattern is more prominent on certain colors. It therefore may be:

Dot Crawl

If others agree then you can search this and other forums for that term and come up with a method of reducing it that is applicable to your source. Because of the cadence of animation (i.e., the number of repeated frames) not all methods may work as well with your source.

[Selur]

Waifu2x (https://github.com/sunnyone/Waifu2xAvisynth) should work fine for older cartoon/anime content.
Since it's there aren't much detail to begin with, most denoisers probably can do the job too if you just crank their strength up.
Even cranking up EZDenoise to high values (15) during deinterlacing with QTGMC should work.

[jagabo]

KNLMeansCL() with some masking based on saturation...

[themaster1]

the video is 760 x488, fix that to begin with i mean recapture

[killerteengohan]

KNLMeansCL() with some masking based on saturation...

WOW!! that looks like it worked wonders.

Did you use default filter settings with no extra parameters?

Did you also do other things to this? It looks like it had more than just KNLMeansCL used on it.

[killerteengohan]

Was that noise available in the tape or caused by bad connection during capturing? I've seen bad TV reception could cause that so most likely recorded that way, There maybe some restoration filters that can eliminate such noise but will affect the sharpness of the video.

I never noticed it when watching it on the TV from the VCR directly. But then again, supposedly the VCR changes the output from the actual tape with processing? or so they said.

It's my friends tapes, and he did the extraction using this. He just wants me to make it look better for him since they have less encoding experience and know I primarily work with animation and CGI video.
https://github.com/oyvindln/vhs-decode/wiki/Diagram-Visuals

His description of it was basically this.

the method we use is the best current method though I can assure it
the process basically bypasses the VCR and taps straight into the VCR's head
it takes that signal, and decodes it with modern software
so analogue to digital conversion then decode through whats basically a software VCR.

basically straight capture of the analogue source without any fuckery from these ancient VCR's with ancient video tech
the VCR is a slave for it's drum to capture the signal on the tape

They have made the bold claim that it is the best possible method for VHS extraction without needing to spend a HUGE amount of money on hardware.

I don't extract VHS tapes, or know much about doing it. I am just fixing up the video and this diamond pattern is an issue I have not seen before with video's.

[jagabo]

Their process isn't separating the chroma carrier from the luma, leaving dot crawl artifacts over areas of high color saturation. On top of that there's the usual noise from VHS tape's low signal to noise ratio.

Here's an improved version of what I did earlier. It retains more detail. You can adjust the KNLMeansCL and SMDegrain parameters to get more or less noise reduction.

LWLibavVideoSource("Yu-Gi-Oh! (1998) Tape 1 (01-03) (VCTM-02203) (VHS-RETAIL-NTSC-J) (Mono) (1).mkv", cache=false, prefer_hw=2) 
AssumeTFF()
ConvertBits(8)
ConvertToYV12(interlaced=true)

BilinearResize(480, height)

SeparateFields()
even = SelectEven().SMDegrain (tr=2, thSAD=500, refinemotion=true, contrasharp=false, PreFilter=4, mode=0, truemotion=true, plane=4, chroma=false)
odd = SelectOdd().SMDegrain (tr=2, thSAD=500, refinemotion=true, contrasharp=false, PreFilter=4, mode=0, truemotion=true, plane=4, chroma=false)
Interleave(even, odd)
AssumeFieldBased()
AssumeTFF()
Weave()

TFM()
TDecimate()

pre = last
KNLMeansCL(d=1, a=2, h=5.0, channels="Y")
smask = SaturationMask().ColorYUV(gain_y=400, off_y=-100).Blur(1.4)
Overlay(pre, last, mask=smask)

Spline36Resize(760, height)