Fixing (not upscaling) old VHS-Camera shot

[Platos]

I have here an old camera-shot which was one of these VHS-cameras (video below).

Has anyone of you an idea, which filters i can use for this?

[Sharc]

I have here an old camera-shot which was one of these VHS-cameras (video below).

Has anyone of you an idea, which filters i can use for this?

How did you capture this? What tools did you use? Your setup?
You should start with a decent capture: lossless YUV 4:2:2, interlaced.

[dellsam34]

Yep, requires a recapture unless the tapes are gone in which case it is better to leave alone, anything you do to it will make it even worse.

[Alwyn]

@Platos, if that MKV isn't the capture file, you can upload the captured AVI up to 500mb on here. That will give the experts more to work with. As is, it's not very good at all; very blocky, wonky colour.

If it is the capture file, you need to do a recapture increase the bitrate a lot (OBS? at least 8000kb/sec) or capture lossless AVI.

The video itself looks to be in pretty good shape so if you crank up the bitrate, you'll end up with a nice donkey.

[Selur]

recapturing with a higher bit rate seems to be the only sensible thing,... (otherwise your best hope are machine learning filters, which try to add some details; attached DPIRDeblock+BasicVSR++)

[Alwyn]

That's pretty good, Selur!

[Selur]

That's pretty good, Selur!

Problem is, that the source is just too compressed and broken,...
(script: https://pastebin.com/7Gc2tNNr)

Cu Selur

[Platos]

Sorry guys, i did not get a notification.

How did you capture this? What tools did you use? Your setup?
You should start with a decent capture: lossless YUV 4:2:2, interlaced.

You mean with capture what camera? I dont know, i did not capture this. This is a camera used in the mid 90' till early 2000.

If you mean how i did digitalize it, then it was one of these video capture cards, where you could use to convert from analog to digital.

Yep, requires a recapture unless the tapes are gone in which case it is better to leave alone, anything you do to it will make it even worse.

They are not gone, but how should i recapture it (higher bitrate? other device ?)?

@Platos, if that MKV isn't the capture file, you can upload the captured AVI up to 500mb on here. That will give the experts more to work with. As is, it's not very good at all; very blocky, wonky colour.

If it is the capture file, you need to do a recapture increase the bitrate a lot (OBS? at least 8000kb/sec) or capture lossless AVI.

The video itself looks to be in pretty good shape so if you crank up the bitrate, you'll end up with a nice donkey.

The Capture file is on a videocassette. So analog tape. I can not have the original file in digital version. But the uploaded file is only a short cut of the full video ( i can not upload here). But it does look exactly the same.

To check the original file is not that easy. I do not own any display which have SCART cable.

recapturing with a higher bit rate seems to be the only sensible thing,... (otherwise your best hope are machine learning filters, which try to add some details; attached DPIRDeblock+BasicVSR++)

I did not digitalize it by myself, so i can not say how they did it. In my memories i do not remember these videos were better. But iu would have to check it. But i would have to buy a SCART to hdmi cable or sth like that to connect a videoplayer to my tv/monitor. There is no other way to find out.

BasicVSR does (as i feared) look very artificial. I will try the DPIRDeblock.

I tried this one (need linux and command line): https://github.com/miccunifi/TAPE

But it did not help that much on 576p. Strangely it looked better, when i used 460p somehow. Maybe this is because of downscaling. (picture comparing below)

You have to watch it fullscreen probably for seeing the difference)

Edit: Your videos are btw. wrong aspect ratio. When i use DPIRDeblock i get also videos with wrong aspect ratio

[dellsam34]

They are not gone, but how should i recapture it (higher bitrate? other device ?)?

Follow along here: https://forum.videohelp.com/threads/414350-Capturing-S-VHS-in-High-Bitrate

[Platos]

ok, thanks. I will search the capturing hardware tomorrow which was used for this and ask here, if this is good enough. Maybe the settings were just bad when they got captured.

[Platos]

They are not gone, but how should i recapture it (higher bitrate? other device ?)?

Follow along here: https://forum.videohelp.com/threads/414350-Capturing-S-VHS-in-High-Bitrate

Ok, i did now search the videos and the capture-system. I found out, there are even 2 capture-devices (dont know why). One is called "elgato video capture" (this one: https://www.digitec.ch/de/s1/product/elgato-video-capture-windows-mac-ios-game-recorder-416828) and the other "Magix retten Sie ihre Videokasetten" (this one(looks like on the thumbnail): https://www.youtube.com/watch?app=desktop&v=YGn5fxgbwfg).

I do not have s-video. I use this scart-cable. The Videoplayer is one of these (JVC HR-J670EU): https://www.ricardo.ch/de/a/vhs-videorecorder-jvc-hr-j-670-1240113674/


I did try both and used VirtualDub2 and this is the result (uploaded Raw-Files (without audio)).

My subjective impression is: They looks both more detail-rich. I feel like i can see the fur of the donkey much more detail-rich. But there are even more aggressiv "lines-artifacts". So my question is: Why is that? My setting where: Pal-B, 25FPS, "no compression UYVY", and "720x576 UYVY YUV 4:2:2 interleaved"

Is this because of the tape (i do not remember it ever looked like that played on a tv over exactly this videoplayer)? Is it because of the videoplayer (maybe i clean it? maybe there is a state of the art device i can buy used?) ? Is it because of the caputure-device? Or is this because of wrong capture-settings?

And most important question: How can i fix this line-artifacts?

[Sharc]

That's already a significant propgress compared to your initial example in post#1.

And most important question: How can i fix this line-artifacts?

I see 3 main issues with your new attempts:

1) Interlacing.
What you see on your PC are the interlaced scanlines which is exactly as it should be for an analog tape capture. This is normal and not an artifact and it will disappear with deinterlacing, either done by your player/TV at playback time, or by postprocessing the .avi capture using a deinterlacing filter (Avisynth, VirtualDub ....)

2) Dotcrawl
You got a lot of dotcrawl (flickering moving dots). This has its origin in the composite video signal coming out of your VCR. It can be avoided by:
- using a better VCR with S-video output instead of the SCART/composite of your model
- using a decent capture card with a good internal "comb filter" instead of your current USB dongles
- using a DVD recorder in passthrough (composite IN -> S-Video OUT, with its comb filter enabled)
- applying a dotcrawl/rainbow filter in post processing

3) Line wiggling/flagging.
- A VCR with TBC, a capture device with TBC functionality, an extarnal TBC or a DVD recorder in passthrough will solve it.

Lastly, you captured into RGB which is just a waste of bitrate for VHS video. Capture into YUV 4:2:2. Configure your codec accordingly.

[Platos]

Thanks for all the advice. I will test the deinterlace. Which one would you recommend for this source? QTGMC?

2) Do you have a sugestion for a specific model i could search for on second-hand platforms ? Maybe when i get a nice price i would think about buying such one (I mean Videoplayer)

3) What is TBC ?

And how can a DVD-Recorder be usefull for video-tapes? I can't put video tapes in it to record.

And about YUV: I could not take that. VirtualDub said not supportet by capturing device. But when i have to deinterlace anyway (so reencode), then i can just take sth else when reencoding. So doesnt matter then.

Edit: And what do you think is a better source for editing? The magix one or the elgato one ? So in case i would not buy another capture device.

Edit 2: I think Hybrid does detect it (wrongly) as progressive. But i have no clue what all these terms say (like telecine, field-blended and so on): What should i take here:

[Sharc]

Thanks for all the advice. I will test the deinterlace. Which one would you recommend for this source? QTGMC?

QTGMC is perfect for very most cases.

2) Do you have a sugestion for a specific model i could search for on second-hand platforms ? Maybe when i get a nice price i would think about buying such one

Look for a model with TBC
https://forum.videohelp.com/threads/286055-VCR-buying-guide-%28S-VHS-D-VHS-Professional%29
https://www.digitalfaq.com/forum/video-restore/1567-vcr-buying-guide.html
https://gleitz.info/forum/index.php?thread/48526-%C3%BCbersicht-%C3%BCber-alle-jvc-pan...r-deutschland/

3) What is TBC ?

Time Base Corrector. Tons of discussions and debates in this forum about TBC.

And how can a DVD-Recorder be usefull for video-tapes?

It stabilizes the picture by providing some basic TBC functionality like "line TBC" preventing line wiggling. Plenty of posts and threads in this forum.
https://forum.videohelp.com/threads/403635-Panasonic-DVD-recorder-passthrough-settings...15-DMR-ES25%29

Edit: And what do you think is a better source for editing? The magix one or the elgato one ?

Neither is convincing. But you may judge yourself.

Edit 2: I think Hybrid does detect it (wrongly) as progressive. But i have no clue what all these terms say (like telecine, field-blended and so on): What should i take here

.avi doesn't have an interlaced flag, so Hybrid (or any other tool) can't know. You have to set it manually. Your captures are interlaced, TFF (=top field first).
For the rest: Learn video basics.

[Platos]

ok, thanks, then i will try QTGMC Topf Field First and lets see.

And i will have look on this list, thanks. Maybe i find a cheap one out there.

[Sharc]

Here an attempt to process the elgato capture, doing something like

AssumeTFF()
QTGMC() #deinterlace
spline36resize(360x576).spline36resize(720x576)  #horizontal downscale-upscale trick for VHS to reduce dotcrawl

[Platos]

Oh Wow, this is nice. I tried it now with QTGMC TopFieldFirst and it still hat all these stuff.

So this is first deinterlace and after that resize width by half and then back to original resolution, right?

Do you know how to do that with selurs hybrid?

[Sharc]

Oh Wow, this is nice. I tried it now with QTGMC TopFieldFirst and it still hat all these stuff.

So this is first deinterlace and after that resize width by half and then back to original resolution, right?

Yes. The downscaling involves lowpass filtering which suppresses the dots but doesn't significantly affect the native poor horizontal resolution of VHS.

[Platos]

Thanks man, you are the hero ^^

Do you know how to do that on selur's hybrid? If not, it is also ok, i can ask on the hybrid thread.

I did manage to do that by 2 steps: deinterlace and resizing to 360x576 and then in a second encode resizing back. But maybe i have to ask in the hybrid thread how to do that in one step.

[Alwyn]

The downscaling involves lowpass filtering which suppresses the dots but doesn't significantly affect the native poor horizontal resolution of VHS.

Clever. Works well in VDub too. Throw in the CCD filter and Donkey looks good.

[Platos]

Clever. Works well in VDub too. Throw in the CCD filter and Donkey looks good.

how can i do that in virtual dub and what is a ccd filter ? could you explain ?

And do you do that afterswards or directly on capture?

[Alwyn]

Yes sorry, I fell for the old trick.

CCD is the Camcorder Colour Denoise filter:

32bit:
https://drive.google.com/file/d/11GkZAfP4wVRBsswi6892H7eLsvxyGjIs/view?usp=drive_link

64bit:
https://drive.google.com/file/d/1aM4dk0_6Kqvi29JabnCTkqOjljzbD06M/view?usp=sharing

Set up the Deinterlace filter to YADIF, Double Frame Rate, TFF

Do the resize as Sharc -described as follows:

Add the Resize filter, set up the first resize (to 360) exactly as per this:


[Attachment 78732 - Click to enlarge]

Hit OK

Now, add another instance of the Resize filter to go back to 720x576:

[Attachment 78733 - Click to enlarge]

Hit OK

Finally, add the CCD filter and set it up like this:


[Attachment 78734 - Click to enlarge]

Hit OK twice.

QTGMC does a better job at deinterlacing.

[Sharc]

In fact one may experiment with the horizontal downscale-upscale to find the compromize between dotcrawl suppression and preservation of the horizontal details/sharpness.
Btw the credits for this trick go to jagabo. As far as I remember I picked it once from one of his scripts.
The line wiggling/flagging is still there of course. Only some sort of line TBC would really eliminate it. I masked the borders a bit to make it appear less annoying (cosmetics).

[Platos]

In fact one may experiment with the horizontal downscal-upscale to find the compromize between dotcrawl suppression abd loss of horizontal details/sharpness.
Btw the credits for this trick go to jagabo. As far as I remember I picked it once from one of his scripts.

Good idea. I will try 480 and 540 on width (720: 1.5=480 and 720: 1.333=480).

And thank you both very much !!

[Platos]

The line wiggling/flagging is still there of course. Only some sort of line TBC would really eliminate it. I masked the borders a bit to make it appear less annoying (cosmetics).

Yes, but i saw they are pretty expenisv (second hand). So for now i will stay with that result.

BUT: can someone explain me why the new capture all are 5:4 instead of 3:4? Resolution is the same (720x576) but aspect ratio isn't ?? How is that possible? Can i somehow change that in virtual dub ?

[Sharc]

The line wiggling/flagging is still there of course. Only some sort of line TBC would really eliminate it. I masked the borders a bit to make it appear less annoying (cosmetics).

Yes, but i saw they are pretty expenisv (second hand). So for now i will stay with that result.

BUT: can someone explain me why the new capture all are 5:4 instead of 3:4? Resolution is the same (720x576) but aspect ratio isn't ?? How is that possible? Can i somehow change that in virtual dub ?

You have to study "Anamorph" picture storage, means the context between Frame Dimensions (720x576 in this case), Display Aspect Ratio (DAR, like 4:3) and Sampling Aspect Ratio (SAR) aka Pixel Aspect Ratio (PAR). It can be quite confusing, I know. Especially because the same 3-letter synonyms for various "Aspect Ratios" like PAR, DAR, SAR, FAR, MAR are used ambiguously.
Simplified explanation for your case: Your captured frame is stored as a frame of 720 width and 576 hight. 720/576=1.25=5:4. That's where the 5:4 comes from. However, this picture is anamorphic distorted, means it is horizontally slightly squashed. For viewing it correctly as 4:3 it has to be horizontally stretched. This stretching is done by the player (TV) reading the picture's SAR aka PAR (which is in very good approximation 12/11 for standard PAL VHS captures like yours), or just forced by the player to play it as DAR=4:3 rather than 5:4. Alternatively, you can stretch it yourself by resizing it by means of a script or a tool (GUI) to represent a 4:3 picture. Confused? No wonder. You have some buzzwords. Tons of posts in this forum on this subject. Google for it and have fun . Maybe start reading here:
https://en.wikipedia.org/wiki/Pixel_aspect_ratio

When you right click into the picture in Vdub you get a list of some standard Pixel Aspect Ratios (PAR, aka SAR). Play with these selections and see and try to understand what happens.

[Selur]

if you want to step deeper into that, read the stuff linked over in https://forum.selur.net/thread-597.html

[Platos]

Thanks to both of you. Ok, i think i understand that somehow. Isnt that somehow the same principle like when i screen-capture my 16:9 monitor while watching a 4:3 video. After the capturing the video still looks like 4:3, but it is 16:9 video then, because the black bars are like "burned" into the video. Or is this another story?

But to the relevant topic: Let's say i want, that the video is always played at the right "stretch", is there no other solution beside changing resolution afterwards?

The strange thing is: The old source (From #1 in this post) is also 720x756 pixel, but it has the right 4:3 aspect ratio. So in this source the pixel aspect ratio is then 1:1 (when i understood correctly before). So i feel like this was made somehow directly on the capture.

So is it maybe possible to make that in virtualdub while capturing? Or why is the other source from #1 same resolution but 4:3 aspect ratio than my capture? I feel like i missed some important settings on capturing.

When you right click into the picture in Vdub you get a list of some standard Pixel Aspect Ratios (PAR, aka SAR). Play with these selections and see and try to understand what happens.

Edit: I saw this right now: When i left or right click in the picture in virtual dub nothing happens. I can see the video playing but i can not click into the picture nor beside it.

Edit2: Ah, you meant in the normal mode, not in capture mode.

[Sharc]

The strange thing is: The old source (From #1 in this post) is also 720x756 pixel, but it has the right 4:3 aspect ratio.

The tool which you used to produce the .mkv flagged the DAR explicitiely as 4:3 in the .mkv container, so the player which reads this DAR flag got instructed to play it as 4:3.
When you play it in Vdub (not forcing the 4:3 playback) the frame is still played as 720 wide and 576 high which is 5:4.
Now right click into the picture and select "4:3 frame (TV)" and you get the 4:3 view as 768x576 (including the small black side borders).

[Platos]

Ok, Update: I found out i can change output PAR in Hybrid and then it is the same resolution, 4:3 aspect ratio (So DAR) and i can change the "stretch" of the Video.

For testing i took 1150x1053 (as described in Selurs link for analog sources), but it is not the same as in source from post #1. But actually im not sure which one of them is the right one. Actually quite difficult to tell.

Is 1150/1053 common for the VHS-8 PAL tapes ?

Edit: I tried also 12/11 and this is less stretched than the source from #1. Difficult to decide

Edit2: Yeah ok, the one source from #1 is 16/15. But i just dont know which stretch is the original one.