VideoHelp Forum
+ Reply to Thread
Page 1 of 8
1 2 3 ... LastLast
Results 1 to 30 of 229
Thread
  1. Hi .. recently had some really good help on correcting a poor VHS capture.
    https://forum.videohelp.com/threads/359358-Post-Capture-help-on-restoration-correction

    Very pleased with the end result, and particularly grateful to Sanlyn who put a lot of time into this, as well as others such as Jagabo & Conucopia, your combined efforts were tremendous.


    Most of my other VHS captures were from my own Camcorder so no multi generation copies, and easier to sort out, been managing to apply some of what I learned here.
    I do have one other 'bought in' video that I need help with, this was my own wedding ... which the family would like me to get onto DVD, it was a 'Professional videographer' and as I have the 'master' tape, it is only one generation from original (or so he told me) .... but it is a 30 yr old VHS recording with all that this brings.


    I have uploaded 2 sample sects ........ would really appreciate any help and advice, and if you can spare the time, explain the steps you take to analyze what needs doing ..... would help with my education, and may also help others.
    You guys carry out feats of Technical excellence, and sometimes I look and have no idea how you figured out what to do, keen to try & learn.
    Image Attached Files
    Quote Quote  
  2. Member DB83's Avatar
    Join Date
    Jul 2007
    Location
    United Kingdom
    Search Comp PM
    Oh my. You did have a fine head of hair back then.

    I may be generalising here but irrespective of the original source you effectively now have a VHS and the usual rules apply. Noise, colour levels etc.

    Also, does the video have indoor scenes in the church, reception etc ?. These would have to be treated differently since the church would rely on natural light whereas the reception could have some artificial light sources.
    Quote Quote  
  3. Still have plenty of hair, but now it has natural highlights
    You are right it has indoor scenes, happy to upload sample.
    Hoping to learn the sequence of what to do to evaluate and how interpret the results, so you know what changes to make.
    Quote Quote  
  4. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    Agreed. The levels in this capture are unworkable: it is too bright, contrast too high; anything brighter than a skin highlight is completely blown away. One might have said that the earlier videos could have problems that were flukes or due to age; the problems here are not only similar but are more severe.

    I fail to understand why fans of DV continue to praise the merits of capturing VHS to DV, whether they use a bitrate set heaven-high or not. Be that as it may, the tape does not appear to be a "master" but looks multi-generational. The combination of player DNR and oversharpening has caused drastic harm, not to mention bad DCT ringing -- neither of which are related to tape aging. The same poor interlacing, apparently from the capture device or setup, appear here as in earlier tapes. Aliasing and sawtooth effects look much worse here.

    I would have to suggest: surely there must be a better capture setup for these tape. Unless, of course, the source tape used here really does look and play as seen here, and the hallowed DV process is indeed an accurate representation of the input. It's a shame that this is seems to be the only copy of the event. Not that it's impossible to improve, but this is a bit like taking a sledge hammer to an automobile before trying to repaint it.

    Aside from that, it's a right proper looking couple and an attractive wedding party. Hopefully there's a way to get a better capture.
    Last edited by sanlyn; 19th Mar 2014 at 06:40.
    Quote Quote  
  5. This was provided 'finished' on a good quality tape, Videographer stated it was first gen master created from original.
    I followed basic advice to get it converted to DV before tapes degrade further.
    I played it in my Panasonic deck (was expensive one at the time) S-Video out, via TBC and ADVC300 direct to FireWire port of PC. Captured using WINDV, no post processing carried out yet.

    Other than setting ADVC300 for PAL, all settings at default.
    I no longer have the TBC or ADVC300 so not really set up for recapture.

    I could buy a new capture card and start again, just a pain, as captured about 30 tapes in total ... I did try to do it right, and bought that kit on advice given on this forum rather than use one of the cheap SCART to USB converters.
    Seems I did not get it right though.
    Last edited by Tafflad; 26th Nov 2013 at 07:19.
    Quote Quote  
  6. Member 2Bdecided's Avatar
    Join Date
    Nov 2007
    Location
    United Kingdom
    Search Comp PM
    Originally Posted by sanlyn View Post
    Agreed. The levels in this capture are unworkable: it is too bright, contrast too high; anything brighter than a skin highlight is completely blown away.
    sanlyn, look at the YUV data. There is not a trace of clipping during capture. It's blown away on the tape, but the full on-tape range is faithfully captured. That "nuclear-blue" wedding dress in the garden peaks at Y=224. There's almost nothing at Y=235, never mind above it (EDIT: except, briefly, the sky, which is as you would expect in those shots with manual exposure control and an operator focusing on the bride+groom - there's no brightness pumping). There is no digital clipping. There is no need to re-capture to get everything in range - everything is already within range, and there's nothing you can do to recover what was never recorded 30 years ago.

    However, the chroma is typical of camcorders of that era - wrong colours and illegal colours all over the place. The latter was never a problem on CRTs, but is a serious problem on PCs.


    If you greyscale() it you'll see the luma levels are surprisingly good and consistent. I don't think you could expect or get any better from a camcorder tape of this vintage, and you have all the data faithfully transferred to the digital domain to tweak at your leisure.


    The aliasing and halos are very strange and quite troubling. I've tried the ideas Didee suggested for my own slightly aliased camcorder sources and they're ineffective - it'll need something stronger.


    Sanlyn, how you attribute any of these problems to DV, I really don't know.

    Tafflad, sorry I don't have any good suggestions for these problems. The aliasing is worthy of a thread on doom9.

    Cheers,
    David.
    Last edited by 2Bdecided; 26th Nov 2013 at 07:13.
    Quote Quote  
  7. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    No, my complaint is about the chroma. Convert this to RGB (as a TV would) and you get some overshoot. But there's something odd about the colorspace (or perhaps it's a little early in the day). Red is depressed, and blue is slightly so. That's not unusual for aged tape. But anyone can try this statement in YUV and observe the results:

    Code:
    ColorYUV(cont_v=90)
    This should increase red contrast and give it a little more spread toward the bright end. But it doesn't. It shrinks red and stretches green. The same thing happens with blue.

    I agree, highlights were blown to pieces a long time ago. The ringing and bad sawtooth edges can be fixed, but a lot of details will slip through the cracks. It's not impossible. Just not something to look forward to.
    Last edited by sanlyn; 19th Mar 2014 at 06:40.
    Quote Quote  
  8. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    Okay, I think I have a working model here. Back in a minute. More coffee.
    Last edited by sanlyn; 19th Mar 2014 at 06:40.
    Quote Quote  
  9. Member 2Bdecided's Avatar
    Join Date
    Nov 2007
    Location
    United Kingdom
    Search Comp PM
    Convert this to RGB and you get illegal colours. CRTs would handle them OK, but PCs will clip them. It needs fixing in AVIsynth in the YUV domain, and I've never found a perfect solution.

    Re-capturing will not make one jot of difference to any of the faults you identify. The problem here is not the capture, though I'd love to know what did introduce the aliasing. If it's just on this tape, what process existed in 1983 that would have introduced it?


    I don't think the problem is the tape ageing either. I bet it looked exactly like this the day it was recorded. I remember a friend's camcorder producing results exactly like this (in terms of levels, colours, white balance etc) 30 years ago when we watched the results at the time.


    Cheers,
    David.
    Quote Quote  
  10. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    The colors and levels can be fixed (well, but the highlights are still gone). I read now where this is the only copy and is not a capture. So be it. That aliasing problem is a real nightmare, and that's the big poohpooh factor. This edge noise and interlacing problem....that's what camcorder transfers look like? I was thinking of trying it for the sake of the VHS->DV fan club. But I think this makes up my mind about it.
    Last edited by sanlyn; 19th Mar 2014 at 06:40.
    Quote Quote  
  11. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    A starting effort, anyway. The script is a mess, will have to clean it up later. QTGMC, jagabo's reduce-width and cleanup for alias and odd line hash, etc., and lots of opposite resizing with nnedi2+sangnom. Color, levels, sharpness, saturation need work too. But I don't know about those edge problems. The mkv is 24FPS PAL interlaced. Making it progressive didn't help (and an odd effect: motion was extremely jerky that way). Time to start surfing doom9 for possible causes of the bad interlacing.

    I don't even wanna think about sample #1.
    Last edited by sanlyn; 19th Mar 2014 at 06:41.
    Quote Quote  
  12. Originally Posted by 2Bdecided View Post


    I don't think the problem is the tape ageing either. I bet it looked exactly like this the day it was recorded. I remember a friend's camcorder producing results exactly like this (in terms of levels, colours, white balance etc) 30 years ago when we watched the results at the time.
    I can remember the camera was a full size 3 tube Sony Broadcast camera .... but I was focussed on other things on the day
    Last edited by Tafflad; 26th Nov 2013 at 11:10.
    Quote Quote  
  13. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    But where did all that distorted motion come from? Surely the original didn't wiggle like that or have severe herringbone effects.
    Last edited by sanlyn; 19th Mar 2014 at 06:41.
    Quote Quote  
  14. Member Skiller's Avatar
    Join Date
    Oct 2013
    Location
    Germany
    Search PM
    No doubt this was shot with a video camera which used a (or three) tube image sensors. Those produce images quite different looking compared to CCD sensors. Most notably there is "phosphor lag" which sanlyn seems to have mistaken for temporal DNR smear. It's not a defect, it's normal.
    Quote Quote  
  15. Hi Sanlyn .. Read some of your other posts on Restoration ... esp on deinterlacing ........ and on many times you say this should not be done if the end result is to produce a DVD for playing on a TV.
    Not in anyway wanting to argue with you ... but would like to understand why you deinterlace in this instance ? or am I misunderstanding the use of QTGMC ?

    The goal here is to produce a DVD.
    Quote Quote  
  16. delete dup
    Last edited by Tafflad; 28th Nov 2013 at 07:20.
    Quote Quote  
  17. Originally Posted by Skiller View Post
    No doubt this was shot with a video camera which used a (or three) tube image sensors. Those produce images quite different looking compared to CCD sensors. Most notably there is "phosphor lag" which sanlyn seems to have mistaken for temporal DNR smear. It's not a defect, it's normal.

    Was def a 3-tube camera, I remember him extolling the version of it at the time (1985) .... and how superior it was.
    Quote Quote  
  18. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    Originally Posted by Tafflad View Post
    Hi Sanlyn .. Read some of your other posts on Restoration ... esp on deinterlacing ........ and on many times you say this should not be done if the end result is to produce a DVD for playing on a TV.
    Not in anyway wanting to argue with you ... but would like to understand why you deinterlace in this instance ? or am I misunderstanding the use of QTGMC ?

    The goal here is to produce a DVD.
    The mkv I posted was re-interlaced, which gave smoother motion on playback.

    True, one wouldn't deinterlace unless one has to do in order to repair something. Many filters are specifically designed for progressive video, and operations such as resizing simply make a mess of interlaced/telecined video. In this case QTGMC has motion compensation repair code that requires deinterlacing. Anti-alias, as well, can't work with interlaced effects. Whether or not to deinterlace depends on the objective and the methods used for cleanup. In the end, interlaced/telecined video output seems to work best for TV. That's not always true, but it's generally the case. A filter like MCTemporalDenosie or NeatVideo can work with either interlaced or progressive, but you'll note that MCTD requires "interlaced=true" for interlaced video and NeatVideo uses a version of SeparateFields in the background for interlaced video.

    It was a little odd that the progressive version was very jerky on my PC monitor. I just didn't have time to look further into that.

    Members here often use mkv for demos in order to preserve storage space. An MPEG is of course what you want for DVD, but it would be a bigger file and would require a bitrate higher than the 5000kbps I used. Camera motion and noise problems always require a higher bitrate.
    Last edited by sanlyn; 19th Mar 2014 at 06:41.
    Quote Quote  
  19. Thanks ... if you are able to come up with any other improvements, all will be appreciated.
    Quote Quote  
  20. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    Originally Posted by Skiller View Post
    No doubt this was shot with a video camera which used a (or three) tube image sensors. Those produce images quite different looking compared to CCD sensors. Most notably there is "phosphor lag" which sanlyn seems to have mistaken for temporal DNR smear. It's not a defect, it's normal.
    Well....OK, I can accept that explanation. I'm still of the opinion that the original video would not have played in such a disturbed manner. If it did, I would certainly have a heated discussion with the photographer. It's a shame that the original isn't available, phosphor lag or not. The DV transfer is apparently all there is to work with.

    The script I posted will hopefully be looked over and, even more hopefully, torn apart for improvements. Admittedly I used elements blatantly stolen from jagabo and some doom9 posts. Smoothing away distortion that actually moves is extremely difficult, even for Avisynth gurus. The resizing and edge cleaning tchniques I've seen so far do generally well, although there is a cost in preserving detail. Some sort of edge masking appears to be the track to follow, but I'm afraid that many masking techniques are my downfall -- so far. Will keep looking around.
    Last edited by sanlyn; 19th Mar 2014 at 06:41.
    Quote Quote  
  21. The original is still available .... well the original I had is ..but as Dave pointed out, does not seem a benefit to be gained by capturing again.
    Quote Quote  
  22. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    I am a fervent critic of VHS->DV capture.To begin with, VHS->DV entails a loss of color data. The captures always look posterized to some degree or other, sometimes very obvious, sometimes subtle (until you get into color correction, and it becomes more obvious). Whatever....I can't see that the original video would be appreciated if played like the transfer, whether on a CRT or not. A particular DV capture device might be "seeing" the original in that way (?). I'd guess that if it were captured in the same way, you'd get the same results. If you look at the video the edge distortion doesn't seem consistent......

    Anyway, I'm still looking. I'm in doom9 now, all the way back to 2002......(!)
    Last edited by sanlyn; 19th Mar 2014 at 06:42.
    Quote Quote  
  23. Member 2Bdecided's Avatar
    Join Date
    Nov 2007
    Location
    United Kingdom
    Search Comp PM
    Originally Posted by sanlyn View Post
    I was thinking of trying it for the sake of the VHS->DV fan club. But I think this makes up my mind about it.
    For goodness sake sanlyn, you're obsessed man!


    There is nothing visibly wrong in this capture that's due to the use of DV, or even the ADVC300.


    I know you think I'm a bit of a DV fanboy, but it's from experience that it's almost never the cause of any problem (trust me, I've pushed it over the years!), and frequently simplifies the solution. By all means use lossless - it won't be worse (unless something goes wrong, or you run out of space) - but please don't misinform people that it'll help them in cases where it won't.


    btw, I wish our wedding video was this good. Yes it has some quality problems, but it's rather well shot for the time.


    Would it convince you if I posted something I'd captured with my ADVC110 from a decent source? Or a lousy source? Or...?


    Cheers,
    David.
    Quote Quote  
  24. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    Well, 2B, it's likely my relative unfamiliarity with digital cams. The last movie camera I used was (believe it or not) an Arriflex 35mm film camera (!!!!). That goes back a way.

    Anyway, rather than undergo torture on the actual taking and showing of the video (nothing can be done about that now, anyway), this is yet another one of those problem-solving learning experiences. The single problem here is that bad aliasing and edge flutter. I've got it down to perhaps 50% corrected. Now looking up some edge masking techniques. As this will have a bearing on one of my own projects, it's for selfish reasons as well.

    I do get your point.
    Last edited by sanlyn; 19th Mar 2014 at 06:42.
    Quote Quote  
  25. Member Skiller's Avatar
    Join Date
    Oct 2013
    Location
    Germany
    Search PM
    I'm almost 100% certain the "aliasing" in this video is not a digital artefact caused by the current transfer to DV or the editing back then but rather an artefact that you get with most tube image sensors because of the way the color is retrieved in such sensor: there is a colored grid in front of the actual phosphor area. This kind of separation is imperfect and as a result you get what looks kind of like aliasing on the horizontal axis. Remember, these cameras had horizontal resolutions in the range 300-400 lines.

    I actually do have a 1984 tube camera right here (why? well, I like old video tech^^) and I get similar "aliasing" if I use it right now, live, without VHS involvement (Composite out -> uncompressed YUY2 capture).


    So, yeah, I think the capture itself is absolutely fine and an accurate representation of what's on the tape.
    Quote Quote  
  26. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    Oh,well. Shucks. As they said in Lawrence Of Arabia,"It is written, then."

    Making progress, but it sure does take a lot of filtering. Keeping at it, though.
    Last edited by sanlyn; 19th Mar 2014 at 06:42.
    Quote Quote  
  27. I don't know anything about tube cameras, but there is a "flash banding" artifact around 12 seconds in the 2nd clip. (if you use separatefields() it's field #599-600)

    It looks very similar to modern day low scan rate CMOS cameras from rolling shutter/partial exposure during photography camera flashes. Is that due to the the "tube camera" as well ? You certainly wouldn't get that with an old CCD design
    Quote Quote  
  28. Banned
    Join Date
    Oct 2004
    Location
    New York, US
    Search Comp PM
    I saw that. I think it's when the photog took a flash camera shot. You can hear the camera click.

    Ahem. Looks I dood it. Smoothed out maybe 90% of the schizo edges. Unfortunately it's really too smooth now (ah, but it's sharp!). Will just have twiddle and tweak now. It's that first half-frame resize step that gets rid of a lot of it, but it kills fine detail. That's the weak spot so far. And SantiagMod, which I've never used before seems like the finishing touch.

    Magic word here is running QTGMC twice. Have to work on that. First run is to deinterlace, second run is with InputType=1 on the progressive clip (I'm surprised it didn't do more damage than it did). It's the motion smoothing and anti-shimmer functions of QTGMC that are doing the grunt work. Guess I'll have to get out the html and see how I can get just those anti-shimmer functions instead of the whole meal.

    Whew! Taking a break.
    Last edited by sanlyn; 19th Mar 2014 at 06:42.
    Quote Quote  
  29. Member Skiller's Avatar
    Join Date
    Oct 2013
    Location
    Germany
    Search PM
    Originally Posted by poisondeathray View Post
    I don't know anything about tube cameras, but there is a "flash banding" artifact around 12 seconds in the 2nd clip. (if you use separatefields() it's field #599-600)
    Is that due to the the "tube camera" as well ?
    I think so, yes. It makes sense. The image is scanned by the electron beam row after row so that would explain it. It is indeed similar to what a modern CMOS would unfortunately produce (I personally prefer CCDs all the way). But to be honest I have never seen the "jello" effect with a tube camera (and I have watched and captured countless hours of shaky family footage recorded with tube cameras), I suppose the beam scans the picture fast enough to avoid that.
    Quote Quote  
  30. If you plan on using QTGMC, I would use border=true. You will see a "blinking" artifact without border=true on the top edge. It's one of the few things that deviate from original TGMC in terms of settings .



    Originally Posted by Skiller View Post
    Originally Posted by poisondeathray View Post
    I don't know anything about tube cameras, but there is a "flash banding" artifact around 12 seconds in the 2nd clip. (if you use separatefields() it's field #599-600)
    Is that due to the the "tube camera" as well ?
    I think so, yes. It makes sense. The image is scanned by the electron beam row after row so that would explain it. It is indeed similar to what a modern CMOS would unfortunately produce (I personally prefer CCDs all the way). But to be honest I have never seen the "jello" effect with a tube camera (and I have watched and captured countless hours of shaky family footage recorded with tube cameras), I suppose the beam scans the picture fast enough to avoid that.
    I was looking for the "jello", but it's not noticable here. Or if there is any, it's pretty much negligible . I expected to see at least some, if "flash banding" was present - unless there is another underlying mechanism for this type of "flash banding" ?
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!