VHS Horizontal Stabilisation

[Keller501]

Hello !

I'm saving a lot of VHS in numeric format at the moment, and last night, I came across a difficulty.
The VHS I just scanned has a terrible stability. The frames moves horizontaly. I've detected a patern of 6 frames that goes to the right and then to the left (3 and 3).
With Avisynth I used the SelectEvery(clip, 6,0) function to see and it seems pretty stable, so the logic now, would be to say, for each frame in that pattern to apply the correct cropping, and then I'd get a correct footage.
Problem, I can't see how I can make a function with Avisynth to treat each frame individualy.
I'm think about exporting each frames as JPG and with a bath cropping each one of them and put them back into a video but.. I guess it could be done with Avisynth.

I hope you understood, thanks in advance !

[DB83]

Numeric format ?

An uploaded sample might assist. Nut you probably just have timing errors which should be fixable using a TBC.

[Keller501]

Numeric, sorry I'm french ^^.

There you go the sample. As you can see only the left/right sides are moving...

[lordsmurf]

Not just any TBC, but specifically line TBC, as found in JVC/Panasonic S-VHS VCRs.

The attached shows a far more serious timing issue, perhaps even capture card rejection by the source. It may need an ES10/15 for strong anti-tearing (this error is similar), as well as a better capture card.

What being used for capture card?

[Keller501]

To capture :

connected to a TOSHIBA RD-XV47

It's the first time I'm seeing this issue.
How do you think I can fix it directly from a re-recording ?

It may need an ES10/15 for strong anti-tearing (this error is similar)

I don't understand that part...

[DB83]

Numeric, sorry I'm french ^^.

There you go the sample. As you can see only the left/right sides are moving...

You might have noticed the right and left sides but the whole image moves.

And as lordsmurf mentions you really need a line-TBC or specifically a brand of dvd-recorder that has pass-through.

No plain re-capture will fix this.

[Keller501]

Numeric, sorry I'm french ^^.

There you go the sample. As you can see only the left/right sides are moving...

You might have noticed the right and left sides but the whole image moves.

And as lordsmurf mentions you really need a line-TBC or specifically a brand of dvd-recorder that has pass-through.

No plain re-capture will fix this.

If you look frame by frame, top and bottom are the exact same. The frame is going right, 3 times and then left 3 times. Every cycle is 6 frames... By just cropping each frame individually the problem will be fixed...
Problem is, with Avisynth I don't know how I can produce a function to do this.
I know I can select every 6 frames and do the crop but it won't be able to treat the frames in between..

[themaster1]

That's what you need : https://www.youtube.com/watch?v=tMA5aH_olAQ&list=FLkdfLPEpmp_VRaJBLnn8iFw&index=109&t=0s

I got one and i love it although i haven't tried with secam signals.

You can find them on ebay or leboncoin for cheap generally (RDR-HXD870,RDR-HXD970,RDR-HXD1070,RDR-HXD890)

[lordsmurf]

To capture :

You're using a junky Easycap. This exact problem is a known issue. The card amplifies timing errors, and looks even worse than it really is.

But still get a TBC, or TBC(ish) at worst.
And upgrade from that $5 Chinese junk card (which you probably massively overpaid for).

Should be an easy fix.

[DB83]

BTW Is this tape SECAM or PAL ?

There could be a problem getting a line-TBC for that.

[Keller501]

It is SECAM.

What card would you recommand ?

[DB83]

Read this topic:

https://forum.videohelp.com/threads/394560-SECAM-L-VHS-TBC-alternatives

Not many capture devices actually support SECAM. I know the Canopus ADVC-300 does (lordsmurf loves these )

[jagabo]

I agree with everyone else: you need a line time base corrector. The cap has lots of duplicate/missing frames at the beginning. That clears up after a while but there's still a lot of trapezoidal distortion, other horizontal wiggles, and different colors on different frames/fields. A good line TBC will clean up most of that. But you can stabilize the existing video with DePanStabilize().

[lordsmurf]

I know the Canopus ADVC-300 does (lordsmurf loves these )

ADVC-300 is craptastic. If you must do DV, at least get the ADVC-100, less screwing with the signal quality. The supposed line TBC does almost nothing in the 300.

ATI AIW does SECAM, of course.
Some others probably do, too, but I rarely have SECAM sources to play with.

[Keller501]

I agree with everyone else: you need a line time base corrector. The cap has lots of duplicate/missing frames at the beginning. That clears up after a while but there's still a lot of trapezoidal distortion, other horizontal wiggles, and different colors on different frames/fields. A good line TBC will clean up most of that. But you can stabilize the existing video with DePanStabilize().

I used this : http://avisynth.nl/index.php/DePan
But can't get it to work..

Read this topic:

https://forum.videohelp.com/threads/394560-SECAM-L-VHS-TBC-alternatives

Not many capture devices actually support SECAM. I know the Canopus ADVC-300 does (lordsmurf loves these )

I'm gonna look into it. But the complexity to find the equipment and the fact that this is the first time I'm seeing this issue (and that I'm just doing it to help, not for professional use), makes me just want to fix this one without buying new stuff Maybe not the right answer

[jagabo]

I agree with everyone else: you need a line time base corrector. The cap has lots of duplicate/missing frames at the beginning. That clears up after a while but there's still a lot of trapezoidal distortion, other horizontal wiggles, and different colors on different frames/fields. A good line TBC will clean up most of that. But you can stabilize the existing video with DePanStabilize().

I used this : http://avisynth.nl/index.php/DePan
But can't get it to work..

After QTGMC() I used:

maxstabH=150
maxstabV=0

vectors = MSuper().MAnalyse(isb = false)
mdata = DePanEstimate(last,trust=1.0,dxmax=maxstabH,dymax=maxstabV)
DePanStabilize(last,data=mdata,dxmax=maxstabH,dymax=maxstabV,rotmax=3.0,method=1,mirror=15,cutoff=0.2,damping=3.0,prev=10,next=10,blur=0)

I also cleaned up the edges of the frame a bit with Crop() and AddBorders().

[Keller501]

I agree with everyone else: you need a line time base corrector. The cap has lots of duplicate/missing frames at the beginning. That clears up after a while but there's still a lot of trapezoidal distortion, other horizontal wiggles, and different colors on different frames/fields. A good line TBC will clean up most of that. But you can stabilize the existing video with DePanStabilize().

I used this : http://avisynth.nl/index.php/DePan
But can't get it to work..

After QTGMC() I used:

maxstabH=150
maxstabV=0

vectors = MSuper().MAnalyse(isb = false)
mdata = DePanEstimate(last,trust=1.0,dxmax=maxstabH,dymax=maxstabV)
DePanStabilize(last,data=mdata,dxmax=maxstabH,dymax=maxstabV,rotmax=3.0,method=1,mirror=15,cutoff=0.2,damping=3.0,prev=10,next=10,blur=0)

I also cleaned up the edges of the frame a bit with Crop() and AddBorders().

Do you have the Depan dll's ? I've only Depan.dll and get an error on the script "Access violation"... I guess i'm missing others dll's

[jagabo]

You need DepanEstimate.dll too.

[lordsmurf]

Trying to fix timing errors with Avisynth, instead of capturing it properly, is really stupid.
And it'll never look anywhere as good, if good at all.

[Keller501]

You need DepanEstimate.dll too.

Okay, found the dll's !

Trying to fix timing errors with Avisynth, instead of capturing it properly, is really stupid.
And it'll never look anywhere as good, if good at all.

Again, I know that, but, it's a really expensive investment, for one VHS, as an individual and also, really hard to find the equipment. So yes, if I had access to that I would recapture. But it isn't...

[johnmeyer]

I don't know if I agree that this is a problem caused by a lack of TBC. Yes, there are a few errors on vertical lines, but I don't see the frame jumping up and down, and certainly don't see any flagging.

There are several clues that no one has yet mentioned which, to me, point to the OP using the wrong settings for the capture hardware/software.

Here are the things no one has yet mentioned:

1. Both Mediainfo and my NLE (Vegas Pro) show this video as progressive. However, a quick inspection shows that it is interlaced. However, whether it is upper or lower field first, I am not sure. I actually had to perform a field switch to get the video to play correctly.

2. Up until around frame 400, the fields are really messed up. After that, the fields are just fine. The initial field reversal issues definitely point to a problem with the capture settings which affect interlacing.

3. The back and forth movement, when viewed field-by-field, rather than frame-by-frame, is quit consistent. This too leads me away from thinking the problem is caused by a lack of TBC because timebase errors tend to have a slightly random movement, and almost certainly not a regular pattern. As the OP noted, the period of the problem is exactly six frames (twelve fields), and is quite predictable. It would be very easy to remove these errors using a motion stabilization plugin, but as others have said, the right thing to do is get the capture to work correctly in the first place.

I do agree that spending the money on a better capture device would be money well spent.

As for the six frames regular pattern, it reminds me a little of some capture devices that attempt to do inverse telecine on the fly. I realize that there is no decimation going on (although the first 400 frames do appear to have some drops), but the regularity of the problem, and the fact that it is close to one of several typical IVTC patterns does make me wonder.

So, my recommendation is to go back to the capture software and look to see how to make the file interlaced. Also make sure all the SECAM/PAL settings for your capture device have been set correctly. Finally, look at the VCR and if it is a PAL/SECAM unit, make sure ALL the settings are correct. Play around a little with whatever options are available and see if you can fix the problem.

BTW, I remember this problem from several years ago in another forum:

Is it possible to stabilize this horizontally jerkyvideo?

The video sample is only 9MB and is still available for download. That OP was using a Datavideo TBC-1000 for the capture, something that is at the other end of the spectrum from what the OP is using.

[Keller501]

I don't know if I agree that this is a problem caused by a lack of TBC. Yes, there are a few errors on vertical lines, but I don't see the frame jumping up and down, and certainly don't see any flagging.

There are several clues that no one has yet mentioned which, to me, point to the OP using the wrong settings for the capture hardware/software.

Here are the things no one has yet mentioned:

1. Both Mediainfo and my NLE (Vegas Pro) show this video as progressive. However, a quick inspection shows that it is interlaced. However, whether it is upper or lower field first, I am not sure. I actually had to perform a field switch to get the video to play correctly.

2. Up until around frame 400, the fields are really messed up. After that, the fields are just fine. The initial field reversal issues definitely point to a problem with the capture settings which affect interlacing.

3. The back and forth movement, when viewed field-by-field, rather than frame-by-frame, is quit consistent. This too leads me away from thinking the problem is caused by a lack of TBC because timebase errors tend to have a slightly random movement, and almost certainly not a regular pattern. As the OP noted, the period of the problem is exactly six frames (twelve fields), and is quite predictable. It would be very easy to remove these errors using a motion stabilization plugin, but as others have said, the right thing to do is get the capture to work correctly in the first place.

I do agree that spending the money on a better capture device would be money well spent.

As for the six frames regular pattern, it reminds me a little of some capture devices that attempt to do inverse telecine on the fly. I realize that there is no decimation going on (although the first 400 frames do appear to have some drops), but the regularity of the problem, and the fact that it is close to one of several typical IVTC patterns does make me wonder.

So, my recommendation is to go back to the capture software and look to see how to make the file interlaced. Also make sure all the SECAM/PAL settings for your capture device have been set correctly. Finally, look at the VCR and if it is a PAL/SECAM unit, make sure ALL the settings are correct. Play around a little with whatever options are available and see if you can fix the problem.

BTW, I remember this problem from several years ago in another forum:

Is it possible to stabilize this horizontally jerkyvideo?

The video sample is only 9MB and is still available for download. That OP was using a Datavideo TBC-1000 for the capture, something that is at the other end of the spectrum from what the OP is using.

Thank you for your help.

About the interlaced content yes I've capture interlaced as progressive i've seen that cutting the sample..
I've checked that it was SECAM in the settings, tried them all (PAL/NTSC/SECAM), and it is secam (as mentionned on the tape)

I'll try another capture as you mentioned. The way I did the capture fixed problems I used to have with audio sync using Virtualdub (even when forcing audio sync)

So, what would be your recommandation, what software, codec etc ?
Was think' of OBS because of nvidia NVENC, I'll see the deinterlace options in the software.

Coming back about TBC and the pattern, as you said, it's always identical and also nothing moves on top and bot of the frame, I presume a bad TBC would cause random movement everywhere and not consistently

[johnmeyer]

Do NOT deinterlace.

Also, do NOT capture as progressive.

Either of those things will screw up the result.

Instead, try to capture the video exactly like it is stored on the tape. The goal of the capture is to obtain a faithful reproduction of the images on that tape, not to manipulate that image while capturing. Once you have it on your hard drive, you can then mess it up as much as you want, but at least you know you have good video to start with.

[Keller501]

Do NOT deinterlace.

Also, do NOT capture as progressive.

Either of those things will screw up the result.

Instead, try to capture the video exactly like it is stored on the tape. The goal of the capture is to obtain a faithful reproduction of the images on that tape, not to manipulate that image while capturing. Once you have it on your hard drive, you can then mess it up as much as you want, but at least you know you have good video to start with.

And as of software, compression and audio sync, what would you recommand ?

[DB83]

TBH I had not considered any of johnmeyer's comments (and he should know since he does this work for a living) since I thought your clip had been re-sampled to .mp4 since these usb devices typically capture to mpeg2.

But what capture software did you use ? Ideally, as stated, interlaced and highest available bitrate. Preferably with a lossless codec - lagarith or huffyv

[johnmeyer]

TBH I had not considered any of johnmeyer's comments (and he should know since he does this work for a living) since I thought your clip had been re-sampled to .mp4 since these usb devices typically capture to mpeg2.

But what capture software did you use ? Ideally, as stated, interlaced and highest available bitrate. Preferably with a lossless codec - lagarith or huffyv

I totally agree.

[themaster1]

Do NOT deinterlace.

Also, do NOT capture as progressive.

Either of those things will screw up the result.

Instead, try to capture the video exactly like it is stored on the tape. The goal of the capture is to obtain a faithful reproduction of the images on that tape, not to manipulate that image while capturing. Once you have it on your hard drive, you can then mess it up as much as you want, but at least you know you have good video to start with.

And as of software, compression and audio sync, what would you recommand ?

Virtualdub to capture and enhance
Codec video: huffyuv (version 2.1.1 not the others!), lagarith OR if you want good compression : UTvideo or H265 (in lossless mode)
Codec: audio: PCM stereo

[Keller501]

Do NOT deinterlace.

Also, do NOT capture as progressive.

Either of those things will screw up the result.

Instead, try to capture the video exactly like it is stored on the tape. The goal of the capture is to obtain a faithful reproduction of the images on that tape, not to manipulate that image while capturing. Once you have it on your hard drive, you can then mess it up as much as you want, but at least you know you have good video to start with.

And as of software, compression and audio sync, what would you recommand ?

Virtualdub to capture and enhance
Codec video: huffyuv (version 2.1.1 not the others!), lagarith OR if you want good compression : UTvideo or H265 (in lossless mode)
Codec: audio: PCM stereo

Perfect, will redo the capture tomorrow. Hopping no audio issue (I used to capture like that, but as soon as the tape had a glitch, the audio was lost in sync)

[jagabo]

If you can't get A/V sync with VirtualDub try AmarecTV instead.

[DB83]

Perfect, will redo the capture tomorrow. Hopping no audio issue (I used to capture like that, but as soon as the tape had a glitch, the audio was lost in sync)

I am sure there is a setting in vdub to handle audio-sync which then drops video frames accordingly.