Some wierd probs with a old NTSC Tape

[Gerry62]

Hi,

Looking at getting the chroma sorted out on this tape a firend asked me too look over. It has short busts where the chroma is all stuffed up.


http://i51.tinypic.com/fdya3a.png

Here is a clip of what it looks like, It also has a second or two of the colour that looks fine. Just so you can see the difference!

Now at the 0.14 mark there are one or two frames that show the 'real' image. Its very short.
My friend has only given me a short sample to play with sofar. (no point is stuffing about with 10 or so Gigs when Its not salvageable)

http://www.megaupload.com/?d=551EG12X

Im only newish to Avsynth and have mainly been sorting out my old Super8 films with it, Now i know it may not be fixable, But hey, Avsynth continues to surprise me every day.
Its also NTSC and I know Pal!

[TheSkiller]

Do you know what machine this was played back with? If these severe periodical chroma issues arise from a very instable signal then a deck with TBC would probably get rid of all these chroma issues. It could be caused by long play mode (or even super long play which is afaik available in NTSC recorders) or by generation loss (tape copied to another tape) or a tape that has been re-recorded many times.

This is the first thing I would look for, if recapturing is not an option then maybe something like this (http://forum.doom9.org/showthread.php?t=161813) is worth a try.


Edit: Megaupload says your file is not available. // Works now.

Edit: It seems no matter how aggressive you filter the chroma, at best you are left with a static rainbow over the picture. The real color information is lost. To see it try TemporalSoften(5,0,255).

[Gerry62]

Thanks for the reply, Just got on to my friend about the sourse material.
Mpeg only copy available, it also exists on tape which has the same probs, only copy to exist. The pattern is 'concrete' into the only known tape copy. Lots of people have 'had a go' to clean this over the last 20 years, all have failed pretty much. I wish he told me that first! crafty beggar he is!


Used some basic filtering.
http://i52.tinypic.com/2ujkob4.jpg


Good Frame unaffected by the pattern.
http://i53.tinypic.com/29b0w7t.jpg

Wonder if a mask with the offending pattern could be then used to tone down the 'bars' a bit, Don't know, Have not used much if any masking before.

[Lyris]

Would it be possible to use frame interpolation on the chroma channels only for the affected frames?

Edit: never mind - just had a look at your clip; I had assumed the bursts of messed up chroma were only a few frames long.

I doubt there is any way to fix that short of treating it as B&W footage and using colourisation techniques... never say never, of course.

[Gerry62]

Would it be possible to use frame interpolation on the chroma channels only for the affected frames?

Edit: never mind - just had a look at your clip; I had assumed the bursts of messed up chroma were only a few frames long.

Sorry should have made that a bit clearer, Hence the tape 'doing the rounds' of 'video experts' for 2 decades or more.

[Didée]

The trick is simply this: there's nothing that could be done.

Bruteforce investigation:

1) Bob().TemporalSoften(7,0,255,255,2)

==> Result#1: The catastrophic color burst are gone. Left over are those thick vertical color bands that can be seen in the previously posted screenshots.


2) Use a customized bandpass to filter out the vertical color bands

==> Result#2: the vertical color bands are gone. The frame now essentially is black/white.


This means that chroma is completely fu**ed-up in the affected frames.
Filtering the chroma artifacts makes no sense, because there simply is nothing 'underneath' that could be restored.


BTW ... aside of chroma, there is horrible frame/field-blending in the luma channel. Those cannot be salvaged either.


End of story: Forget about it. A waste of time, nothing else.

[Gerry62]

Thanks Didee, as I thought, Nothing can be done.
Apparently (ive just been informed) the BBC have spent some years trying to fix it, they got close recently then postponed the DVD release. They were going to use B&W film recordings for the luma and then use the Chroma from these rubbish tapes, It's all that now exists.

The recording was PAL in 1970 then converted to NTSC (badly) sometime in the mid 70's to sell to the yanks, some fan recorded it and that's where the colour tapes come from, The B&W kines are quite good, so combining the two makes sence. this has been done a few times with domestic colour over good rescanned Kines. But this tape is affected by rainbowing, about a quater of the 3hr production is OK no rainbowing, it's just the bits like this one. What was done in the 90's was to use the chroma when it was 'good' over the kines then fade chrome off whenever the rainbowing was too much.

Yes the frame/field blending was always a problem with BBC productions. sometimes the OB film stuff is telecined right, sometimes not. The extra PAL-NTSC only makes the issue worse!

Thanks for looking into it, Had tried all sorts, as I like to play for a few hours before seeking help, don't lean anything by getting help all the time!

back to looking into a kine that has a paused frame every 20 or so frames prob is, Its on both fields of the Tape I was given, but not at the same time, so movement and dirt\scratches stay 'locked'. Discard one field and it looks strange halting every second, keep both fields and blend looks better. Still it's a challenge to clean up, which means a lot more to learn!

[johnmeyer]

I get some relief using the ancient CNR plugin:

loadPlugin("c:\Program Files\AviSynth 2.5\plugins\CNR\Cnr2.dll")

source=AVISource("e:\frameserver.avi")
assumetff(source)

a = separatefields()
even = a.SelectEven().Cnr2("oxx",8,35,255,100,255,32,255,false)
odd = a.SelectOdd().Cnr2("oxx",8,35,255,100,255,32,255,false)
c=Interleave(even, odd)
d=weave(c)

return d
#stackhorizontal (source,d)

Here's a poor (low bitrate, therefore somewhat blocky) encode of the result:

Chroma Reduced Result (http://dl.dropbox.com/u/1561578/test.avi)

Be aware that when using settings as high as I used, you will definitely get some chroma lag on fast moving, highly saturated objects.

Play around with the settings and perhaps you can get even more relief. This is a port of a filter originally done for VirtualDub, so you could instead use it in VD which lets you more interactively play around with the settings (you'll have to remember to add before/after "separatefields() equivalents in VD and put the CNR filter in the middle).

[2Bdecided]

It's this one, isn't it?
http://archive.whoniversity.co.uk/av_faq/av_faq.html#3.1

...and this is the delayed release...
http://wipednews.com/2011/01/05/the-ambassadors-of-death-recolourised/
...in which case it sounds like they're going to use this...
http://en.wikipedia.org/wiki/Colour_recovery#From_chroma_crawl
...which should work far better than anything you can get from those tapes!

Cheers,
David.

[Gerry62]

Yes it is!

But what you have to remember from the colour recovery, It only works on TR's that the Chroma was left on. Ok so how many prints are like that, well sadly very few..
If it was possible, the Mind of Evil (71) would be fully colour by now.


Just read the article of wiped new and AOD can be restored, thats a good thing then, but why not Mind of evil, again, no chroma on those at all.


Ive been on a crusade to find missing TV/Radio for 25 years!
you may have seen lots of my work/finds!

[jmac698]

Crazy, ppl in those 20 years never tried hard enough.
I know there is color there, and I have an idea to recover it. Of course you won't believe until shown, neither will I - but now I can prove the color *is* there.
Here. You can see the image peeping through.

DirectShowSource("E:\project001a\sampleproblem3\Issues NTSC.mpg")
assumetff
yuv2yhs
yhs_viewamp#view saturation (this is not true saturation)

#(Extract from) YHS Tools by jmac v0.3
#Requires Masktools v2 and YV12 video
#Functions to convert to and from YHS which is Y, Hue, Saturation - almost
#Simply finds the phase and amplitude of the YCbCr values

function yuv2yhs(clip y){
y
dist="x 128 - 2 ^ y 128 - 2 ^ + .5 ^"#sqrt(x^2+y^2)
deg="180 pi / * 360 + 360 %"#(*180/pi+360) % 360
sc1="224 120 / * 16 +"#scale 120 to 16-235 for saturation
sc2="224 360 / * 16 +"#scale 360 degrees to 16-240 to stuff in U
amp=mt_lutxy(utoy,vtoy,dist+" "+sc1)#close to saturation, max=120 (I think)
#ph is close to hue but starting at blue-magenta
ph=mt_lutxy(utoy,vtoy,"y 128 = 0 "+dist+" x 128 - - y 128 - / atan 2 * "+deg+" ? "+sc2)
ytouv(ph,amp,y)
}

function yhs_viewamp(clip y){
vtoy(y)
}

function yhs_viewph(clip y){
utoy(y)
}

[Mounir]

What's the purpose of this script jmac i don't get it or do use you other tools for this analysis to be meaningful ?
that's what i get with your script:
http://imageupload.org/?d=C6B6BB8C1

[Gerry62]

I get the image with Monochrome(ish) bars. I expect its the colour. Don't know. Interesting little script though

[um3k]

It's a greyscale representation of the color saturation. It shows that there are correlations between the color and the luma, which implies a better than a zero chance of recovering the original colors. I took a shot at doing so the other day, but my limited mathematical knowledge precluded successful recovery.

[2Bdecided]

It's a greyscale representation of the color saturation. It shows that there are correlations between the color and the luma, which implies a better than a zero chance of recovering the original colors.The saturation is encoded by the chroma amplitude. The actual colour (red or green or blue etc) is encoded with the chroma phase. It's the phase that's wrecked AFAICT.

Cheers,
David.

[Gerry62]

Wrecked is the word for it!
The Beeb, I think, are going to manually colour those bits and use the real colour (the odd frame or two) as reference, but don't quote me. etc...