Well, VHS capturing Mostly cartoons, having some questions:)

[zerowalker]

So true, and here it is.

it´s much TBC issues if you look at the door frame, this is what stabilizes much with QTGMC, but i need to retain interlacing as it´s truly 50i
1 lines per frame (i think).

The newest thing i tried is:

QTGMC( Preset="placebo",sourcematch=1,EdiThreads=4,noisep rocess=0)
SeparateFields().SelectEvery(4,0,3).Weave().Assume TFF()

in the end, not to sure if that´s any good though, going from interlaced to prog to end up interlaced again;S

[jagabo]

QTGMC() and the other filter sequence aren't meant to correct VHS jitter.

[zerowalker]

See that you haven´t got the overlay problem stated anymore, was it wrong?

And well i know that they aren´t really there like an alternative TBC, but QTGMC really smoothed that heavy motion.
But if there is another way to get what i want, i am all in for it.

As currently i have to go with what i said before, dfttest (for denoise). then QTGMC(sourcematch=1) for Smoother motion or what to call it.
And well it´s Very very slow, about 1-2 fps with 4ghz Quad, MTmode(2,5) (5=100% cpu for me).

So if there is any suggestions on the script, or any fault in it, as i pretty much copied it from another please tell me

Thanks

[sanlyn]

I had no problem with QTGMC on the shimmer_raw clip. The wiggly lines can't be helped much with software. But I do see excess pixel activity on camera motion, mostly fixed with QTGMC and some low MCTemporalDenoise. The last of the motion gnats died after a touch of NeatVideo used only asa a temporal filter, other filter parms turned almost off. Edge jaggies and edge ghosting were obvious. Couldn't fix all of that with the time available tonight. The color balance is fairly awful, likely from tape deterioration: there's a green color cast, darks are crushed, blue is oversaturated. I fixed some interlace and edge jaggies quickly, but I finally gave up on discoloration, chroma bleed, and cyan staining. I keep cruising doom9 for solutions for those.

Frame 47 of the original: in the entire image there's not a single clean black, gray, or white. The "real" colors are anyone's guess. Highlights were destroyed long ago, no way to recover them. The image looks oversaturated, with rather grim darks.

[Attachment 12686 - Click to enlarge]


After some ColorYUV, I used a VirtualDub gradation curve filter derived from loading this clip into the Color Finesse plugin in AfterEffects (still trying to get a firm handle on that app!). I'm not satisfied with colors yet, but levels are more TV-friendly, gamma still looks a bit strange, the gal keeps turning yellow and mama stays a dark purple-gray. But the woodwork looks like real wood. Anyone's guess what color the time/date characters are, but I don't think they're gray or white.

[Attachment 12687 - Click to enlarge]

Histogram from the YUY2 original. I can't say I've seen luma clipped at the left in quite this this way. First time for everything.

[Attachment 12688 - Click to enlarge]

The attached mpg is PAL 25fps interlaced. I used QTGMC (slow), then MCTemporalDenoise (Low, edge clean + anti-alias on), some very light NeatVideo at nearly temporal-only (I think I should have turned NV's Y-channel entirely off, 20% still looks over-filtered). Motion looks smooth to me. But a line-tbc and a better player would have solved a lot of problems.

[jagabo]

See that you haven´t got the overlay problem stated anymore, was it wrong?

I have no idea what you're talking about.

[themaster1]

This is the best i could do with your footage zerowalker: http://www.mediafire.com/?a6m31xcwgljfdll

[zerowalker]

Hmm, weird it should be more understaturated then over, atleast from my tests and comparisions against some things that are released on dvd.
But remember that it´s PAL, you have to remove every other chroma line, else it will be messed upp with converttoyv12(interlaced=true).

Yeah QTGMC helps alot, but i don´t know how to use it, without going to progressive and then back to interlaced;S.
Colors are well, it´s an old tape, and it´s fully interlaced, so not much to expect, a good vhs player can probably get the best out of it.

and Jagabo, what i meant was:

. The nnedi3, motion interpolation, dfttest is getting confused by the date time overlay which isn't moving with the picture. That's what's causing the odd motions.

And themaster1, that looks pretty nice, it seems fully interlaced as well, can you show the script?

[themaster1]

Sure thing download it from here: http://www.mediafire.com/?3mzcnxvcamohb4r
I think it need more tweaking and i suggest you work with rgbeq and cmill a little more.

[jagabo]

and Jagabo, what i meant was:

. The nnedi3, motion interpolation, dfttest is getting confused by the date time overlay which isn't moving with the picture. That's what's causing the odd motions.

Ah, I had forgotten I originally wrote that. Yes, I was seeing some odd motions around the date/time stamp. I removed it from the post because I decided that wasn't what you were talking about.

[zerowalker]

themaster1, that was one hell of a complicated script;O
don´t really know what i can learn from it, but thx for sharing!

jagabo,
oh, well the overlay date is a bit fuzzy and stuff, but if the dft and qtgmc is getting confused by it, i will probably need a solution for that;S

[sanlyn]

Hmm, weird it should be more understaturated then over, atleast from my tests and comparisions against some things that are released on dvd.

I think that's mostly a matter of color preference. I don't see that this color can be cleaned up without some serious masking and RGB work. The only major color work I did was to try to reduce the blue and cyan casts, especially in the brights. But it would take much more work.
[QUOTE=zerowalker;2166878]But remember that it´s PAL, you have to remove every other chroma line, else it will be messed upp with converttoyv12(interlaced=true).
I don't know what that means. NTSC YUY2->YV12 works the same way.

Yeah QTGMC helps al ot, but i don´t know how to use it, without going to progressive and then back to interlaced

It's done all the time. I used this:

## ------------------------- ##
##   ColorYUV omitted here.  ##
##       Input is YUY2
## ------------------------- ##

FixVHSOversharp(30,14,10) 
ConvertToYV12(interlaced=true)
AssumeTFF().QTGMC(preset="slow")

MCTemporalDenoise(settings="low",edgeclean=true,ECmode="dehalo_alpha()",\
  enhance=true,AA=true,useEEDI2=false)

# -- reinterlace -- 
AssumeTFF().SeparateFields().SelectEvery(4,0,3).Weave()

# -- more cleanup for aliasing and dot crawl --
daa().daa()

# -- restore 4:2:2 for further color work --
# -- If encoding with HCenc, ignore this statement. --
# -- Normally I avoid this, but it Worked OK here.  --
ConvertToYUY2(interlaced=false)

# -- crop bottom border noise and center image --
Crop(0,0,0,-8).AddBorders(0,4,0,4)

# -- use for RGB, NeatVideo, and TMPGenc only. --
# -- Otherwise, ignore the statement below.    --
ConvertToRGB32(matrix="Rec601",interlaced=false)

Colors are well, it´s an old tape, and it´s fully interlaced, so not much to expect, a good vhs player can probably get the best out of it.

What does interlaced have to do with degraded color? That's caused by tape aging and improper storage.

The nnedi3, motion interpolation, dfttest is getting confused by the date time overlay which isn't moving with the picture. That's what's causing the odd motions.

What "odd motions?". Playback is unstable; tape almost always requires a line tbc, even with smoother playback. Edges are still distorted, they're simply blurred by dfftest. Any temporal smoother would have the same effects. You have to be more specific about the "odd motion" description. The motion in this clip is camera motion. I don't see the date/time characters moving around (but I do see dot crawl and aliasing, there and in other areas). I don'[t see that nnedi3 or dfttest are confused. I do see a lot of blurring. IMHO dfttest is mostly for degrain and seems to be set too strong here. Try something like TTempSmooth, or if you think grain is a major problem here, try TemporalDegrain.

themaster1, that looks pretty nice, it seems fully interlaced as well, can you show the script?

It's a handy script that I've used before with good results. It converts back and forth between YV12 and RGB, but that's common practise in many complex scripts. Completely avoiding YV12 eliminates far too many valuable filters; YV12 is often necessary for many types of color correction, masking, and motion interpolation that can't be done in any other way. Simply over-smoothing video won't solve many problems with damaged analog sources.

Here are 2x bicubic enlargements of the date/time characters from the three sample clips: mine, the master1, and your processed avi -- but not shown in that order. Cleaning up detail like this would require more work, but these are initial efforts representing three approaches to processing. I'll leave up to readers to determine which image is from which sample.


[Attachment 12697 - Click to enlarge]

[zerowalker]

What i mean is, PAL is "YV12" in YUY2 form or something like that, so to convert it properly you have to do this.

AssumeTFF()
Mergechroma(SeparateFields().Blur(0,1).Weave())

Then you can converttoyv12 (the TFF part is not essential for the convert though).

it´s jagabo who said about the motion interpolation, i don´t know myself so can+t say anthing there, but indeed
dfttest does blur, but i do compare with the original to see how much, and with these settings i think it´s not bluring any real details.

Oh didn´t know that, but i really like the look on the middle picture, as with my settings it get´s like in the first one, it´s like you are watching through a window or something.

Oh, thought Fully Interlaced would degree all kinds of quality, as it´s very low details remaining, and half chroma with PAL.
But well, when i look at it, it seems pretty good on colors, except White and black, white just blurs and like fades into stuff, while Black isn´t black, more like grainy smudge or something, so if colors are dark, the smudge.

And what i meant with proper vcr, like real vcr machine, those that broadcast use, though to get a hand on those, you will probably need to work at such a place, or know such people, which i sadly don´t;S

But i have seen the work of those on old tapes, and it´s in a whole different league, can´t really be compared at all.

[sanlyn]

What i mean is, PAL is "YV12" in YUY2 form or something like that, so to convert it properly you have to do this.

AssumeTFF()
Mergechroma(SeparateFields().Blur(0,1).Weave())

Then you can converttoyv12 (the TFF part is not essential for the convert though).

I've seen that script used, but recently in connection with using QTGMC in YUY2 instead of YV12 -- as was explained back in February in the QTGMC thread on doom9. Perhaps other members can explain the use of that script for usual YUY2->YV12 conversion and whatnot, but I haven't seen it used in the normal conversion process. PAL and NTSC VHS are encoded in a form of YUV that is YCbCr - not exactly YV12, but something along similar lines, even if the two systems don't treat color in exactly the same way. The reason why most people capture PAL/NTSC VHS as YUY2 is because it's more "like" the way the tape is played back. Your capture is YUY2. The MergeChroma statement is used to restore only the chroma processing, not the Y-data. I don't see that anything material is gained or lost by using it that way. With damaged video there will be further processing anyway.

it´s jagabo who said about the motion interpolation, i don´t know myself so can+t say anthing there, but indeed
dfttest does blur, but i do compare with the original to see how much, and with these settings i think it´s not bluring any real details.

Actually, it's blurring everything. But if you find it suitable, that's up to you. I'd use other methods.

Oh didn´t know that, but i really like the look on the middle picture, as with my settings it get´s like in the first one, it´s like you are watching through a window or something.

The top image is from your processed AVI.

Oh, thought Fully Interlaced would degree all kinds of quality, as it´s very low details remaining, and half chroma with PAL.
But well, when i look at it, it seems pretty good on colors, except White and black, white just blurs and like fades into stuff, while Black isn´t black, more like grainy smudge or something, so if colors are dark, the smudge.

The smudge and vague detail is the result of colors and luma crushed during capture. The brights are terrific problems, partly due to the cam era lighting, to the use of mixed light sources, and the use of forward firing flood lights that has many of the same effects as direct flash. The auto-exposure meters in consumer cameras do more harm than good with that type of lighting, and it's not easy to fix. I could have done more work, but time is against me right now. Tape aging didn't help.

Deinterlaced/reinterlaced PAL or NTSC isn't "half chroma". QTGMC was designed to avoid that and other interlace problems. The video I submitted is re-interlaced, as the script shows.

And what i meant with proper vcr, like real vcr machine, those that broadcast use, though to get a hand on those, you will probably need to work at such a place, or know such people, which i sadly don´t.

Pro-grade studio VCR's are nice, but who can afford them? The VCR's I use now are Panasonics from the 1995-1996 era, and the better ones have very smooth playback. I found them on eBay. When new, they ranged from $250 - $400 USD retail, but I got them used at a fraction of that cost. True, I had one or two cheap bad buys with VCR's that were over-used or not cared for, but I managed to get 3 in excellent condition. They're used in conjunction with DVD recorders used as line-tbc pass-thru devices. There are numerous posts in this forum that discuss using old DVD recorders for tbc pass-thru. I would not recommend using them to record directly to DVD with old or damaged sources. Once the dirt and noise gets encoded to digital form in a DVD recorder, it's almost impossible to clean. Capture to PC, clean first, then encode.

Tapes that have been replayed many times will not load smoothly back to the feed reel, so they don't feed off those reels very smoothly during repeated playback. An old VCR trick is to repack the tape -- that is, you fast-forward the tape without playing or pause entirely onto the takup reel, then rewind without pause or playing back to the feed reel. You might have to repeat this a time or two, the idea being to more smoothly rewind the tape onto the feed side without all the bumps and dips in the winding. It does help feed tape into the playback path more smoothly.

[zerowalker]

Not sure what you mean, but if it´s processing YUY2 directly, it´s of course better, but i don´t think it really matters, and yes they the NTSC and PAL are different in the way that PAL don´t use every other chroma line (something with using one line in two frames or something i think), while NTSC uses both.
That´s why i have to blur it away, as if i convert to YV12 interlaced, it will change color as every other chroma line is red or green, which will result in a video shifting color every frame.

oh, well you see, ultra smooth, not to my liking really, but don´t know how to keep it sharp.

Yeah did some extra checks, and idd it blurs to much, it seems fairly good at sigma=5, and noiseprocess=0 on qtgmc.
If you know other ways i am all ear, mctemporaldenoise in your script, didn´t do to well, don´t really like it as it´s terribly slow on the good parts, but it´s good many options so i guess you can make some neat stuff out of it.

Yeah indeed, sad that the VHS is the most common recording storage from before 2000 or something, and there playback machines doesn´t even handle them perfectly, as the tapes are able to contain very noise details, but the playback devices can´t show it in it´s original form.

Well will get a good one sooner or later somehow, but currently can´t get my hands on any, and yeah read about the line tbc through dvd, seems a neat option, but there wasn´t so much information on it, i think there was something with difference on PAL and NTSC devices there aswell, not sure.
And yeah, agree totally, encoding on the fly on all forms are bad, especially mpeg-2 which only perform good on very high bitrate and 2 passes.

Oh, nice trick, never had any problems with unsmooth tape playing, but good to know if i run into any

[jagabo]

What i mean is, PAL is "YV12" in YUY2 form or something like that, so to convert it properly you have to do this.

AssumeTFF()
Mergechroma(SeparateFields().Blur(0,1).Weave())

Then you can converttoyv12 (the TFF part is not essential for the convert though).

I've seen that script used, but recently in connection with using QTGMC in YUY2 instead of YV12 -- as was explained back in February in the QTGMC thread on doom9. Perhaps other members can explain the use of that script for usual YUY2->YV12 conversion and whatnot, but I haven't seen it used in the normal conversion process.

It's to eliminate hanover bars in PAL soures.
http://en.wikipedia.org/wiki/Hanover_bars

[sanlyn]

There are many used PAL DVD recorders, Panasonic and Toshiba from 2000-2005 are most numerous. You connect the VCR output to the DVD recorder, then connect the DVD's output to your capture device. Nothing else to set. Their tbc circuits aren't adjustable. Just play the tape "through" the DVD, the same way you'd watch tv thru them without recording.

With very few exceptions, all of the better Avisynth and VirtualDub filters are slow. With damaged video source, good work simply takes time.

[jagabo]

PAL and NTSC VHS are encoded in a form of YUV that is YCbCr - not exactly YV12

Analog NTSC is closer to YUY2 -- every scanline has unique chroma channels. Analog PAL is closer to YV12 where pairs of scanlines share chroma channels (the AL part of PAL).

[zerowalker]

Yeah, good explanation there, i wasn´t too informative, sorry for that.

Yeah but wasn´t there some problems with color degradation and stuff from some, and there was only a very few tested?

[sanlyn]

Analog NTSC is closer to YUY2 -- every scanline has unique chroma channels. Analog PAL is closer to YV12 where pairs of scanlines share chroma channels (the AL part of PAL).

Yes, thanks for clarifying. So, now, I have no PAL sources, but with PAL tape would one capture directly to YV12? Or YUY2?

[jagabo]

Yes, thanks for clarifying. So, now, I have no PAL sources, but with PAL tape would one capture directly to YV12? Or YUY2?

It depends on how the capture device (or drivers, or software) treats the chroma channels. I would try it both ways.

[sanlyn]

Yeah, good explanation there, i wasn´t too informative, sorry for that.

Yeah but wasn´t there some problems with color degradation and stuff from some, and there was only a very few tested?

If you mean DVD recorders used as pass-thru: They all do some inner processing in that regard, but I believe Panasonic gets some of the bad press about messing around with color and levels. My ES20 doesn't seem to cause any of those problems, and usually these can be fixed with software or a proc amp. My preference has always been Toshiba, which has given me the best and cleanest results. You'd be amazed at how much cleaner the output can be with a line tbc.

[jagabo]

A line TBC is essential if you plan on making DVDs. All that time base jitter eats up bitrate. And obviously, the video looks better when it's not jiggling around all the time.

https://forum.videohelp.com/threads/306272-Computer-video-capture-vs-vcr-to-dvd-combo?p...=1#post1882662

[zerowalker]

YEah well i really want it, and will get one, but when i do, i want to get a good one, not a half good one, as i have already made one misstake;S
But will try to keep an eye on the DVD Passthrough on the secondhand market here, ebay etc;D


Yes so true, but currently making them as, well more of a backup for the ones that needs the video (as you saw it´s an old recording).
But will keep the tape and do a better cleaner jobb when i can get a hold of more professional equipment, cause i really want to make the best out of these old tapes, as there is only one example, it´s not like a movie that are available around the world.


EDIT:

Can someone, if possible, upload a clip of maybe atleast 10 sec, one with line-tbc one without?
I have seen pictures and such, but would be nice just to see real samples, if it´s possible.

[themaster1]

I think i have found a good compromise for this scene zerowalker:

picture (original,previous tweaks, new tweaks):
http://www.mediafire.com/?wsi3kkn92vg2g6k

vdub config (include filters if needed):
http://www.mediafire.com/?utyggnpad283i18

video:
http://www.mediafire.com/?0vg7d6a6cl3cwb9

[jagabo]

Can someone, if possible, upload a clip of maybe atleast 10 sec, one with line-tbc one without?

The animated GIF linked to in my earlier post should give you an idea.



That's a 4x nearest neighbor enlargement slowed down to 4 (?) fps. On the left is the straight from the VCR, on the right is with the Panasonic ES15 passthrough.

Various VCRs with/without TBC:
https://forum.videohelp.com/threads/230650-Confused-Why-a-VCR-with-TBC-if-separate-TBC-needed-anyway

[zerowalker]

themaster1, your script is really complex, but it gives very nice resulst, though i don´t know how to tweak it, but the functions are the ones that really does some neat work, sadly i can´t get the NLMeansCL to work (AMD GPU).

But i got one problem with your script in comparasion to dfttest, your gives ghosting, or it´s bad at motion estimation i think, and i don´t know how to tweak this.


jagabo, yes i remember that one, i understand what it´s doing and even with just that information there i know that it will be a huge difference with 25 fps all over the place.

does the ES15 have any problems?
as i remember one needed some amp to get the right colors or something.


is NV-HS 900 any good?
or does it have some denoising stuff going on?

Cause i Maybe can get my hands on one, if it´s worth it.


Look at the red ring:

First = themaster1
Second = Mine Dfttest,motioncompensated + qtgmc

Edit: My bad, was using Santiag for the Zigzag dot crawl, which breaks interlacing.

[themaster1]

Can't you upload the video sample for this scene ?

[jagabo]

does the ES15 have any problems?
as i remember one needed some amp to get the right colors or something.

There's a very small levels and color shift. Easily fixed since your going to be filtering anyway.

https://forum.videohelp.com/threads/319420-Who-uses-a-DVD-recorder-as-a-line-TBC-and-wh...=1#post1982320
https://forum.videohelp.com/threads/319420-Who-uses-a-DVD-recorder-as-a-line-TBC-and-wh...=1#post1983288
https://forum.videohelp.com/threads/319420-Who-uses-a-DVD-recorder-as-a-line-TBC-and-wh...=1#post1982737

[sanlyn]

If in decent condition, the NV-HS900 is a very good machine.

I don't see ghosting in either scene. Each image has its own good points, and with a tweak or two themaster1's script would produce results even more similar to #2. Both have edge ringing or halos. but that's not ghosting, which loo0ks almost as its name suggests: it's an offset, usually half=-transparent double image.. The red ring in the top image looks the same as that in the lower one, just not as much sharpening (that's easy to fix). Both have the same chroma smears (not easy!).

How does Santiag "break" interlacing? It just doesn't work well on interlaced video.

ghost example, from really awful DVD transfer. Note double image to right of figures and to the right of white diagonal lines:

[Attachment 12705 - Click to enlarge]


Fixed . . . well, almost. Not 100%. Detail is so poor to begin with, stronger filter would destroy more.

[Attachment 12706 - Click to enlarge]

[sanlyn]

never had any problems with unsmooth tape playing, but good to know if i run into any

zerowalker, for shame. Where do think those wiggling lines come from that we see in all your posts?

Can someone, if possible, upload a clip of maybe atleast 10 sec, one with line-tbc one without?
I have seen pictures and such, but would be nice just to see real samples, if it´s possible.

Jagabo posted excellent links and samples earlier -- some of them I keep losing track of, I can never find 'em when I want them. In an earlier thread from you I posted a number of clips that addressed smoothing playback with a pass-thru, as well as copy protection and tape damage problems. There are many video samples that are much longer than 10 seconds.

https://forum.videohelp.com/threads/331681-s-video-artifacts?p=2141386&viewfull=1#post2141386
https://forum.videohelp.com/threads/331681-s-video-artifacts?p=2141384&viewfull=1#post2141384

In another thread, Orsetto posted some webcam samples. Why, oh why, can I never find these links when I want them?

Except for the posts I mention above, I haven't capped VHS without a tbc since 2003. Those captures are long gone. Look into that Panasonic VCR you mentioned. It just might do the trick well enough if it's in halfway decent shape.