VHS Family Wedding tape restoration

[sanlyn]

Needs work. Will have to backtrack on this one and tweak like the devil. Laters.

Add these steps to the first script I posted (and take LSFMod out of it and leave the output de-interlaced):
QTGMC(preset="faster",TR2=2,InputType=1)
SantiagMod()
LsharpAAF()
GradFun2dbMod(mask=false)
AddGrainC(1.5,1.5)
AssumeFrameBased()
AssumeTFF().SeparateFields().SelectEvery(4,0,3).We ave()

[Tafflad]

Hi Sanlyn ...tyring to run your script ... 2 Q's

Where do I get "Hqdering.avs" all of the usual sites don't have it for download (interestingly they list it - but no file)

Are the 3 "import" lines because those plugins not in your normal plugin folder ? ... and if they are in my folder I just remove those lines ?

[sanlyn]

The text of HQDering is at the bottom of the text file of scripts posted earlier. It requires deen.dll (attached). I made a few changes anyway.

The "Import" lines are loading .avs scripts, which don't load automatically. The only auto-load plugins are .avsi and .dll. I keep QTGMC in my plugins as an .avs script, not an .avsi (why load it and 10 other support plugins automatically every time I run Avisynth?).

Wedding 2: Still get a somewhat soft image because of all the problems being filtered. The attached has a little more clarity.
Wedding 1: You'll never get rid of all the disturbances without destroying it. It had other problems anyway.
Too much re-sharpening after filtering gives bad clay-face effects: the original videos show signs of posterization to begin with.

Scripts are similar, but not identical. Will have to get them together tomorrow. Rain+wind coming from offshore tonight; the cable company's transformers cut off several times (some are still under repair since hurricane Sandy), internet keeps going down. May as well wait and ride this thing out until the A.M. .....

[Tafflad]

Added in the deen.dll.... but still unclear about Hqdeering plugin

In your script you import - Import(ppath+"Hqdering.avs")
So presumably I should have a plugin within plugin folder - Hqdering.avsi
I have no such plugin, and can't find it ...
I did look back through earlier replies in this thraed and can't see it as an attached file ..... am I missing something basic here ?

[2Bdecided]

.avs and .avsi files are just text files that include scripts and/or functions. .avsi files autoload (if in your plugin folder), .avs files do not.

I don't know where Sanlyn grabbed his version from - I know of...
http://web.archive.org/web/20111103213225/http://mf.creations.nl/avs/functions/HQDering-v0.1.avs
and
https://forum.videohelp.com/threads/354272-HQDering-filter


Sanlyn's efforts are quite amazing on this one. I still think the result looks a bit like plastic, but it's less like plastic than you'd expect, given the source. It has an EDI look to it (you get it from nnedi2, eedi2, and anything that uses them like qtgmc) which I've never managed to avoid when using those functions heavily. A usual solution would be to add some of the original noise back in, but you don't want any of that aliasing back so it's hard to know how to do that.

Cheers,
David.

[sanlyn]

Th version of HQdering I'm using is from a 2006 edit by the author, here: http://forum.doom9.org/showthread.php?p=793930#post793930

The "plastic" look already shows signs in the original sample. Posterization during resizing makes it look worse, so I added some grain. Also tried resizing using the dither() plugin, but it actually looked worse. One could add a little more grain, I think, but usually people who don't realize its purpose get very confused because they feel that any grain at all is incongruous with "digital" video, even if the source is analog or film.

Every anti-alias and edge cleaner I can find performs some sort of resizing, either up or own. The step that did the most damage to fine detail in the badly aliased Wedding2 was the downsize to 360x576 and up again. If you perform that but do no processing while the vid is in that state, it comes back to 720x576 fairly intact and with much less of an edge ghosting or double-line effect -- but with the bad aliasing pretty much unchanged. The only real "denoiser" used was QTGMC. If you specify QTGMC in "yadif" mode, it's not quite as plastic but, alas, it's really blurry. Use a typical sharpener (like LSFMod or its predecessors), and it's Plastic City. It would appear that the sampled 2nd-gen copies of both vids had some processing during the copy operation -- oversharpening, some oversaturation, temporal DNR, and a contrast bump. That might have been the VCR or camera playing the original, or the capture device. (Also noted: despite the fact that luma levels in the capture are at least fairly well controlled, there simply are no highlights left at all).

As for HQdering....I don't think it accomplished that much. There is some edge ringing, but the worst damage looks more like right-edge ghosting that has been sharpened. Dering and dehalo filters had little effect on that. I actually made some progress with the ancient FixVHSOverSharp, but unfortunately in only works in YUY2, and YV12->YUY2->YV12 came back with a dried-up, posterized look.

Another problem, was the odd DV capture colorspace, where contrast commands on chroma channels worked exactly the opposite from standard. Each camera shot in both sample captures had the usual whacko VHS fluctations in luma and color balance. If you bump saturation a little in one shot, the next shot looks oversaturated and the third shot looks washed-out. If you lower contrast in one overly contrasty shot, the next shot looks thin and one-dimensional. Add some red to one shot, the next shot is too red., etc. The later mkv's 1 & 2 illustrate these problems: they use a few similar filters, but there are major differences between the two scripts....which I hope I can edit down from a huge mess into something readable to post.

None of these scripts are gospel. Surely they can be improved. There are plenty of old pros around here who know much more than I do, especially when it comes to masking techniques.Hmm, and 2BDecid's note on the "EDI look" was most interesting. Definitely something I'll take a look at.

[2Bdecided]

"EDI" look: Compare qtmgc with dumb bob (no other processing - just compare those two double-rate deinterlacers). Don't look at still images, just watch the resulting video in real time. Through all the noise, line-twitter and other faults (which your eyes will see past after a few minutes), bob looks more like a degraded record of something real, qtgmc looks more like someone re-created the original event as an oil painting. The less real detail there is in the original, the worse it is.

I don't know if other people notice it, but it bugs me. If you like qtgmc, and don't see this fault, it's probably better if you don't try the above comparison on this clip, because once you've trained yourself to see it it'll probably bug you too - and without qtgmc, what are you going to do? There isn't an objectively better deinterlacer.






I wasn't criticising your result sanlyn. I'm impressed with it. I couldn't have created nearly such as good a result as you have, and like I said, the plastic look is far less than I'd expected.


Cheers,
David.

[sanlyn]

I have no problem seeing what you decribed. Have always seen those effects, and their severity depends on the input as you noted.

[mammo1789]

I think high value cameracolordenoise vdub will remove almost all of the "bluegreenish" on the hat on the women

[Tafflad]

I have no problem seeing what you described. Have seen those effects often. Their severity depends on the input as you noted. They appear to be unavoidable with these samples; to a visible extent they exist in the originals.

Do you want a sample from later in tape for comparison?

[sanlyn]

I think high value cameracolordenoise vdub will remove almost all of the "bluegreenish" on the hat on the women

I tried it. Well.....it didn't accomplsih much. That color is really corrupt, mostly reflections from all the vegetation and no sunlight to provide relief. The photog should have used a polarizing filter to ward off the reflected green and cyan overcast. Too late for that now, I guess. I have no idea what was done with the original master to make the DV copy, but.....Oh, well.

[sanlyn]

Do you want a sample from later in tape for comparison?

You might try that later. Meanwhile, I changed the uploads in post #33 and made some notes, plus scripts.

[Tafflad]

Having looked at the clips some 200 times, I concluded that the dresses are various shades of green, from soft bright to very pale, with the bride in white + pale green (?) dress; and the hats have a dark green cloth base attaching them to the ladies' hair. I think.

Hi ... just looked again and I think I know what you are describing ... on Sample 2 ..
( Just had the wife to look as well - as the foremost authority on the dresses )


Frame 447 for example .... the bridesmaids dresses are aquamarine .... the flower garlands in the hair were made of flowers alternate between one white & one in aquamarine colour ... slightly darker than dress ... but nothing much. There was no dark green material underneath ... the flowers were attached direct to a band .. no green.
The Brides wedding dress was all while - no green anywhere.

I'll load a clip that shows colours clearer.

I can scan in one of the wedding stills if you want an accurate colour reference.

[sanlyn]

Thanks for that info. Well, then, if the bride wore all white, the chances that it will actually look white from that video are a long 'way off. That blue discoloration is on captures from this and other threads. At least we have the aquamarine fairly well correct on the other dresses. Indded there are also flashes of dark green "speckle" that appear in odd locations. Those could be tape damage.

Can you tell us again, what did you use for capture? I tried looking for that in other posts but can't find the info. Capture device? Player. ? It's not unusual for discoloration to appear on old tapes. But it is unusual when it all looks exactly the same way in all scenes on different tapes and affects only the same colors.

[Tafflad]

Loaded a clip - wedding sample 3 .... couple of snipped together sections where I think colour of Brides and Bridesmaids dresses is clear .... opening sequence shows the flower garlands in the hair - and the alternate flower colours.
Hope this helps.

[Tafflad]

Can you tell us again, what did you use for capture? .

High Quality VHS Tape (PAL), played in a Panasonic NVS HS800 (SVHS 3 head deck) .. output SVideo into AVT-8710 TBC, then SVideo again, along with L&R audio feeds into Canopus ADVC-300.
Output of this Firewire into PC (no capture card) and used WINDV to capture files.

I know you passed a comment previoulsy on SONY codec, I did not specifically load or use any Codec (I didn't realise I had to), but it seems it used Sony DVSD

I purchased the ADVC-300 and AT-8710 on advice on this forum to get the optimum out of my tapes. Appreciate it may not be everyones answer, but when I asked for help that is what was recommended.

[sanlyn]

Sample3 has similar problems. I wonder why no one has noticed that skin tones are orange (definitely the color of nice, ripe pumpkin at the height of the season. I understand that medically this is a sign of severe and fatal liver disorder). There is no facial skin texture at all; colors are so saturated that they bleed over fine detail and obscure it. The colors are seriously unnatural, the DNR is very aggressive. And severe aliasing on bright edges. That capture setup has serious problems. VHS shouldn't look that way, not even when captured to DV.

[Tafflad]

Well as its 30 yrs on, and neither myself, bride or bridesmaids have died of liver disease, must concur that your medical diagnostic abilities do not match your skills with video.

[sanlyn]

Yes, it's encouraging that it's just a bad color problem, and not the "real thing". Unfortunately, not even tape aging can produce effects that are that consistent from tape to tape. It's odd enough that a "pro" photog would use autogain in variable-light situations and end up with the "pumping" seen in these videos -- yet I'm having a difficult time trying to acccept that someone would pay out good coin of the realm for results that some members here have said looks "correct". But I don't own the equipment (except for the AVT-8710, which I've used for years with MacroVision tapes, and which can sometimes be infamous for its dot-crawl), so I wouldn't know how to turn off the special effects in the other components. On the AVT-8710 everything can be set to neutral with the re-set button.

[Tafflad]

All the settings were 'base' on the ADVC-300 other than setting it to PAL i.e. no correction to levels or colour, or sharpening.
Just checked on VHS deck ..
Edit - off
sharp - soft control left at default centre position.
AI - off
S-VHS - on
SVideo - on
Colour -on
Nicam - on

[DB83]

@Sanlyn

Maybe I over-exaggerate but I think you also do. You must rememeber that the OP comes from a part of the world that has as much sun in a year that NY has in a month and that, for the OP, is X times more than I see where I reside.

Out comes the sun and it's straight to the beach. Sun block, even if it were available back then, was never sold since supply would out-strip demand. End result we all end up resembling Miss Piggy

Personally I do not think the skin tones look quite as bad as you make them out to be. But this simply could be down to how we usually see them.

PS I am in love with that blonde maid-of-honour. Is she still available

[sanlyn]

PS I am in love with that blonde maid-of-honour. Is she still available

After 30 years? I don't think so. She wears a wedding or engagement ring in sample #3.

I can't accept the sunburn theory, interesting though it might be.

Having spent 4 summers in the UK, I don't recall 2nd-degree sunburn being a problem there in July or August, not even during the insufferable heat wave of 2005. That color looks more like the Popsicle orange you'd get from cheap indoor tanning lotions. However, I can accept what most experienced videographers would describe as color corruption and chroma loss going from analog VHS (YCbCr) to 4:2:0 DV-AVI using a capture device and codec designed for DV->DV transfer (which is not the same thing as VHS->lossless capture, nor even VHS->MPEG2 capture), and which itself entails a double round of lossy encoding if you want anything other than DV playback. I won't go into the Canopus ADVC debate, either, especially with the Pinnacle/PowerDirector crowd -- and especially that this setup appears to be all that is available to the owner...for now, anyway. I've seen too many samples of VHS->DV capture over the years to find it to be an advantageous way to digitize and archive VHS tape.

Sample3 does look less damaged in some respects, but it would still be a chore -- both because of tape aging and because of the method of capture. I would definitely not discard the tapes.

And BTW, even in sample 3 you can see the level problems inherent with VHS. The first camera shot looks washed out (in fact, it is washed out) and has a definite soft-focus look (intentional?). Filter settings for that camera shot won't work for the later scenes.

[Tafflad]

@Sanlyn

Maybe I over-exaggerate but PS I am in love with that blonde maid-of-honour. Is she still available

That is my sister, she has been married twice since this video was taken.

[sanlyn]

Ahhh, thank you. My wife and I were both correct!

Well, I'm not taking sides here with the ladies. The sister indeed has that look of an Irsh corrine (although I think your standard corinne usually has darker hair), quite photogenic -- but having "been there" with photogenic people in my wild youth, male and female, I can vouch that after the cameras stop rolling things usually look somewhat different. Be that as it may, I find the bride and groom alike to be equally interesting. And I'm having trouble with the hair color on a couple of the other ladies. This is all academic and wildly impractical for most viewers, part of learning to work with color layers and ColorFinesse in After Effects.

After a little color correction with #3, the flowers on the bridesmaid's' heads look pink and white, the bride's look white (but they are usually blown away with no detail, so it's difficult to say). The hair color and texture of the older man with the bride (her father?) looks rather odd as well. Will have to tackle this later. It's the weekend here, and wives don't understand that video work should have priority over shopping for food......

[Tafflad]

Irish .... my God Sir, there be no Oirish in my family.
Welsh, with a smattering of English ....

OK trying to help with some more information.
Attached sample 4 .. which is indoors, so different lighting .... so no UV issues as mentioned for outdoors.

(Contrary to views of DB83 we were not suffering from UV deprivation, and there was no perma-tan involved )

If it helps to have a non-video colour reference ... I attach a hi quality scan of a photo from the day taken on 2 1/4" film, no digital cameras at that time.
Had to zip it as forum does not accept TIFF file format. - scan0001

[DB83]

My friend. Sarcasm is the lowest form of wit. And I was laughing more at myself than you

Since I understand that recapture is not an option. If it might help. Maybe you could ask that nice man who captured that NTSC tape for you.

[sanlyn]

Ah, yes. Welsh. Now that does seem more fitting.

[manono]

High Quality VHS Tape (PAL), played in a Panasonic NVS HS800 (SVHS 3 head deck) .. output SVideo into AVT-8710 TBC, then SVideo again, along with L&R audio feeds into Canopus ADVC-300.
Output of this Firewire into PC (no capture card) and used WINDV to capture files.

My capture chain is nearly identical to yours with 2 exceptions. The VCR is an NV-HS860 and I hardly ever use the AVT-8710 box. I use it only if necessary, like correcting a vertical jitter or a tape with macrovision. It does soften the picture and shouldn't be used unless for some particular reason. I also try to get the gamma/levels settings within reason using the software included with the ADVC-300 by capping short pieces and studying them before doing the real capture, and then go over the full capture later scene-by-scene, frame-by-frame, to fix things that aren't the way I want them.

[Tafflad]

OK .. then I got it wrong ....
After asking various Q's here .. I bought the ADVC300 and AVT-8710, and I captured every tape using the same default settings.

In my ignorance and no experience, I had assumed, that I was capturing the 'clean' digtal version, and everything that needed to be changed could be changed once I had the DV file.

I sold the TBC & Canopus units after I captured every tape.

I still have all the original tapes ... I still have the Pansonic deck.
So I 'could' capture again ...but would need to buy another ADVC300 or a PC capture card.
I do have a "Hauppauge wintv pvr 150" digital video recorder card ... but never used that so unsure if that would be any use for capture. (or if W7 drivers exist)
review is here: http://reviews.cnet.com/tv-tuner-adapters/hauppauge-wintv-pvr-150/4505-9332_7-31205053.html

However the earlier comments in this thread seem to suggest capture was OK ... so investing in new capture kit may not improve anything.

[Tafflad]

Ah, yes. Welsh. Now that does seem more fitting.

... I thought you might have said "that explains a lot"