VHS Restoration

[Asesinato]

San >>>Brightness levels goes from -128 to 128

[Asesinato]

jag > You says the trim is different how do i find out what to set it at then?

[jagabo]

jag > You says the trim is different how do i find out what to set it at then?

Trim(x,0) removes x frames from the start of the video. I just start two instances of VirtualDub, open a video in each, find an easily identifiable frame, and note the different frame numbers. Or sometimes I'll use StackHorizontal(v1, v2) to see the two videos side by side in one instance of VirtualDub instead. Ie, if a particular frame is at 100 in one clip and 125 in a second clip, then you need to Trim(25,0) to the second clip to make it sync with the first.

[Asesinato]

Ohh now i see... Great

But ill do all that when me and san has the proc amp settings the way he want them at

[Asesinato]

San >>>Brightness levels goes from -128 to 128

San > Did u see this post??

[jagabo]

Ohh now i see... Great

But ill do all that when me and san has the proc amp settings the way he want them at

Yes, of course. Get your caps right first, then worry about filtering.

You should look into methods of reducing that diagonal line noise though. Make sure the VCR, computer, and capture device are all plugged into the same power strip. If any of the devices doesn't have a keyed power connector plug it in both ways to see if there's a difference in the noise. If you have cable TV coax plugged into any of the devices disconnect it. Remove any other unnecessary cables and devices from all the equipment. If you have different A/V cables, try them. If you have an internal capture card try moving it to another slot if possible -- as far away from other cards as possible.

[Asesinato]

They are all connectet to the same power unit, but so i screen, printer and so on and so on, is that a problem?

I dont have a tv coax anywhere, only the svhs stick from video to computer

[Asesinato]

Ohh and the capture card is in the only free pci slot

[jagabo]

It sounds like you've tried pretty much everything you can to get rid of the diagonal line noise.

[DB83]

Didn't I read early on in this thread that the diagonal noise is only present during this transfer - ie not on a commercial tape.

Meaning that it is already burnt on to the VHS

[jagabo]

Didn't I read early on in this thread that the diagonal noise is only present during this transfer - ie not on a commercial tape.

Yes, he said that.

Meaning that it is already burnt on to the VHS

But it's not burnt onto the VHS tape because it's different each time he plays the tape.

[DB83]

Well we have had 3 or more captures all with different settings. Does the OP mean this ?. Could the DNR on the vcr have some influence in this ? Has that been turned off ?

[sanlyn]

San >>>Brightness levels goes from -128 to 128

San > Did u see this post??

Yes, and thanks for posting. If you used brightness 55, drop it to half of that, or about 25. The low end on a histogram with brightness at 55 was about RGB 40, which is too high. Using just half that brightness ought to do it. In the brighter daylight scenes, try 12 to 15. When you see that the shadows in the dark clothing and the shrubbery no longer look jet black or blocked up, you're on the right track.

The diagonals look about the same, but they do look different in different parts of the tape. In the daytime dinner scene the lines are 5 pixels thick, 5 pixels apart, about 35 degrees. In the dancing scene they're 2 pixels wide, 2 pixels apart, slanted more at 65 degrees. They are about the same in each respective scene.

You can try repacking your tape. This means to fast-forward (without playing) all the way to the end, then fast-rewind all the way back. That helps level the tape inside the reel and can sometimes solve a number of tape flow problems. I don't know that it will reduce the diagonals, but you can try. Poisondeathray's DeStripe() script posted earlier does clean those lines, but it softens quite a bit sort of like a strong temporal filter. I tried other plugins; they had no effect. NeatVideo's temporal function cleared quite a lot of it, but DeStripe caught almost all. There's another similar script from mp4guy on doom9 I've been trying to figure out (Who, me?). I sorta know how it works, but I don't know how anyone came up with coded details.

DeStripe uses a temporal smoother. I feel I could replace it with something weaker. Will give that a try later.

[Asesinato]

New cap with brightness 25.

cap 2 is with DNR on
cap 3 is with DNR off

[sanlyn]

Good job, asesinato. The levels are nearly right-on, don't change a thing. You can even see missing details in the dancing gal's black slacks that were completely invisible in earlier captures. The brightness range 25 to 27 is about right.

This video is rather soft on detail to begin with, so dnr didn't help much. Even with dnr on, the diagonals are still highly visible -- so we'll have to use DeStripe or its equivalent anyway, which will soften detail even more. I think dnr should stay off. Others might differ, but even with it on there's more denoising to do. May as wsell try to keep all the detail we can.

I have a PC repair customer this evening. Back later.

[Asesinato]

Okay here is new cap from bright area with a brightness off 13...

[sanlyn]

Daylight scene, brightness at 13: Very Good! It does make you wonder why the daylight shots need less boost than the darker ones ? ? I've seen it explained that when tape ages and fades, the weaker signals (i.e, the darker ones) show the worst effects. Magnetic tape never has very clean shadows anyway, so something like a night scene would look terrible under these conditions, maybe completely unusable.

The daytime dinner scene with the light streaming through the big window -- that might call for an adjustment somewhere between the church scene and the party scenes. But Try that later. -> Not at my home now, I'm on a customer's PC. Guess I'd better get back to work!

[2Bdecided]

It sounds like you've tried pretty much everything you can to get rid of the diagonal line noise.

It's got to be something physically/electrically/algorithmically close to the VCR, cable, or capture card.

Try switching everything off that you don't need when capturing. That includes the display! Try disconnecting everything that you don't need too.

As a test, try disconnecting the audio too. Just a thought.

Cheers,
David.

P.S. don't wear this tape out!

[sanlyn]

The O.P. states that the noise pattern doesn't appear on other tapes. I've seen these diagonals on other old tapes. In the worst example, the noise occurred in slowly moving waves from left to right. The lines on all these problem tapes were of similar width and frequency seen here. I've read that it's caused by improper tape storage near other magnetic fields (like TV CRT tubes, subwoofers, etc.) or near heaters or in overly warm conditions (such as keeping a tape for several months in the hot trunk of a car parked in bright sunlight). These conditions alter the tape's magnetic layer. There are other signs of poor storage here, such as color fading unevenly.

It's also seen on tapes recorded off antenna or cable; the interference in those cases is in the original signal transmission. Sometimes they look like floating waves, but often they appear as fine-mesh crosshatching sometimes called "FM hash".

I've searched Doom9 English and German forums for a couple of years now. Found occasional posts about this problem. Now and then you find an Avisynth fix for it -- but the fix is always custom-designed for that particular specimen. They tend to use MaskTools and its convolution and lut functions, and temporal smoothers of one kind or another. The scripts don't work on other sources. The authors of these scripts never explain how they work or how they can be customized. There are a few Avisynth plugins around aimed at these problems,, but comments by users say they either do nothing, or ruin the image, or won't run at all (which has been my experience as well). Some of these haven't been updated and require older versions of mvTools, MaskTools, and/or FFT3D. Furnished examples of how they work on sample images don't even work on the images provided.

Thre Dess

[DB83]

I have just noticed that these latest PAL captures are at 29.97 fps AGAIN. This really does not help the cause.

And I still do not detect any sound.

[sanlyn]

Yes, they're 29.97 fps. A "slight adjustment" required in the capture software ? ?
No, there's no audio. Not needed for these tests.

[2Bdecided]

The O.P. states that the noise pattern doesn't appear on other tapes.

I know, but Sanlyn, it's difficult for the interference to be on the tape, but different every time the tape is played. Possible, but it's not the only explanation.

Whereas if the signal on the tape is weak but clean, the FM AGC in the VCR will be cranked up, maximising any interference caused by e.g. having a CRT TV sat right on top of the VCR itself.

So it's possible for the interference to only be visible with this tape, but not actually be on this tape. Other ways this could be possible too (e.g. it's there all the time, but the OP doesn't notice lower but still present interference visible with other tapes)

Given the different noise each time, I'd say chances are 50/50. Given that removing the noise post capture is a pain (or required multiple captures - how much longer is this ancient tape going to survive repeated playing!), I reckon it's worth one last check to try to remove sources of interference while capturing. Might not help, but worth a try.

Cheers,
David.

[DB83]

Yes, they're 29.97 fps. A "slight adjustment" required in the capture software ? ?
No, there's no audio. Not needed for these tests.

You miss the point.

I agree that audio is not important for these tests but I have yet to see a capture by the OP, other than the .ts samples that actually has sound. If he can not capture sound in this way then we are all wasting our time.

A minor adjustment required. Yes, but an important one all the same. All tests should be done with the best available material. Extra frames generated in the capture impair judgement.

[Asesinato]

I can fully capture sound (As startet earlyere to), and will do when needet, but i cant see the reason to do it in capture to make proc amp settings like we want them

[jagabo]

Extra frames generated in the capture impair judgement.

There are no extra frames in his caps. The frame rate is just set wrong in the AVI header.

[DB83]

I can fully capture sound (As startet earlyere to), and will do when needet, but i cant see the reason to do it in capture to make proc amp settings like we want them

That is fair enough. I thought that sound capture settings were standard and one would have to deliberatly turn them off. Do that and it is just so easy to forget to turn them back on.

[DB83]

Extra frames generated in the capture impair judgement.

There are no extra frames in his caps. The frame rate is just set wrong in the AVI header.

Doesn't the avi header determine playback speed ?

But if the avi header is incorrectly written then the capture software is surely suspect.

[jagabo]

Extra frames generated in the capture impair judgement.

There are no extra frames in his caps. The frame rate is just set wrong in the AVI header.

Doesn't the avi header determine playback speed ?

But if the avi header is incorrectly written then the capture software is surely suspect.

This has already been addressed. He can change the frame rate in the header to 25 fps and modify the audio skew a bit to get perfect A/V sync.

But I agree, the software he is using, or the way he is using it, is suspect. But I don't think it has anything to do with the diagonal line noise, bad colors, etc.

[Asesinato]

Jag > I was thinking about the merged, there is something i dont under stand...

v1=AviSource("capture2(1).avi").Trim(97,0) v2=AviSource("capture2(2).avi").ColorYUV(off_y=-24)

That i understand... But what then if you have 4 ? which v should it match up in frames?

v1=AviSource("tt.avi").Trim(18,0)
v2=AviSource("tt2f.avi").ColorYUV(off_y=-24)
v3=AviSource("tt3f.avi").Trim(8,0)
v4=AviSource("tt4f.avi").ColorYUV(off_y=-24) Merge(Merge(v1, v2), Merge (V3, v4))

Trim 18 is the differens from tt 1 and 2
and trim 8 is the differens from 3 and 4

Is that correctly understood?

[jagabo]

I have a hypothesis about the cause of the diagonal line noise. As you may know, VHS records the two fields with an azimuth difference between the two heads to reduce crosstalk from adjacent fields:

http://en.wikipedia.org/wiki/Helical_scan#Slant_azimuth_recording

I suspect the diagonal line noise happens when the head is off center along the track and picking up a little signal from the adjacent tracks. When it is perfectly centered there is no line noise. But if it strays off center a little signal from the prior or next field is leaking in. The reason the noise would occur on some tapes more than others is because some devices overlap the tracks a little more. So the width of the clean center portion of the track will vary depending on the recording device.

This would mean that if you captured the video several times we would see only two different phases of the noise in any particular field (along with noiseless portions). That's a way of testing this hypothesis.