Complex repair of a low-quality VHS/BETA('68) source. How?

[kenpachi]

I'm quite a beginner when it comes to decide on some complicated AviSynth parameters, also I don't feel strong in smoothing video sources. I can do this and that but I'm not experienced very much - this is why I post. I have a low-quality source. It is an NTSC TV broadcast from '68 delivered on DVD. It seems to be kinda VHS/BETA rip but I'm only guessing. I don't know what type of video cameras they were using those days and how they were doing archives.

This is the video:
Video
ID : 224 (0xE0)
Format : MPEG Video
Format version : Version 2
Format profile : Main@Main
Format settings, Matrix : Default
Duration : 18mn 49s
Bit rate mode : Variable
Bit rate : 6 854 Kbps
Nominal bit rate : 7 700 Kbps
Width : 720 pixels
Height : 480 pixels
Display aspect ratio : 4:3
Frame rate : 29.970 fps
Standard : NTSC
Colorimetry : 4:2:0
Scan type : Interlaced
Scan order : Bottom Field First
Bits/(Pixel*Frame) : 0.662

And here is an example:
http://www.speedyshare.com/files/22586894/example.zip

Personally I would try something like (ColorYUV, sharpen, AddGrain, TemporalSoften, ...) and then encode it with 4.1@High, AQ(1.3), PS-RD(1.25) to increase the quality but I'm sure there is much more to try. I don't know which plugins and what values to use. I would appreciate some help, especially a complex script to the source. It would help me on further scripts, no doubt.

PS. The source is interlaced with BFF (MediaInfo and Avs AutoAssume) but when choosing it in Yadif, the output contains duplicated frames each several ones. Choosing TFF doesn't generate any duplicates. That's a riddle.

[Guest]

What is the title and region of the DVD?

[kenpachi]

I think this info can be much help. I've just realized there is an explanation how the source was obtained.

http://img85.imageshack.us/img85/549/simongarfunkelthekraftm.jpg

[zilog jones]

The 16 mm film would have been a kinescope (http://en.wikipedia.org/wiki/Kinescope) film, a common way of recording television back then due to the cost and availability of video tape recording (it's possible this was a live kinescope and was never recorded directly on video tape). This is a pretty messy process (the flickering horizontal banding is one of the artefacts created by this process), especially with NTSC video as they somehow have to get 60 fields of video per second on to 24 frames of film. The addition of (probably mechanical) telecine to U-matic then VHS have added even more noise to each field (as opposed to just the film frames), and the DVD transfer was apparently not done with a TBC (time-base corrector) so a lot of time-base errors remain (horizontal wobbling of video lines) - unfortunately there is not really any way to correct the latter once it has been transferred into the digital domain.

I guess the DVD was encoded using the wrong field order for this source hence de-interlacing only working properly with forced TFF - if this is the case it must look horrible on a TV :/

[Tommy B.]

There is not much you can do about that bad quality, but here, try this :)

source=mpeg2source("example.d2v")



source
checkmate(100)
greyscale()


assumetff()
TempGaussMC_beta2mod()
selectodd()



fft3dfilter(sigma=0.5, bt=3, dehalo=1)

smoothlevels(35,1,255,0,255)
crop(6,2,-8,-10)
spline36resize(720,544)


limitedsharpenfaster(strength=75)


gradfun2db(1.3)
grainfactory3(3,4,2,45,45,45,1.05,1.05,1.05, temp_avg=25)


stackhorizontal(source.assumetff().tdeint().crop(6,2,-8,-10).spline36resize(720,544),last)

You'll need

checkmate
TempGaussMC_beta2mod
fft3dfilter
smoothlevels
limitedsharpenfaster
gradfun2db
grainfactory3

[kenpachi]

@Tommy: Quite amazing if I can say so. Where can I found beta2mod? I looked for it but managed to get only beta1mod. Shame I get only 0.5fps with 2core ;] I wonder if I could succeed with same results using just straight ColorYUV. I'd try the script on my desktop 4core but win7x64 got some problems with the needed plugins.

[Nightshiver]

beta2: http://www.mediafire.com/?bd15iqyizny

I've found no such thing as the "beta2mod". There's beta2u, but no mod. beta2 requires VerticalCleaner.dll (http://home.arcor.de/kassandro/prerelease/VerticalCleaner.rar) With the speed issue, you won't get much faster. TempGauss is inherently slow, no matter if you have 2 cores or 4.

[Tommy B.]

You can use beta1 - the cleaning and interpolation effect is present in both of them.

@Nightshiver
My avs file says:
(version: something point something, beta-1)

Maybe that is beta2u and I have called it wrong...

[Nightshiver]

Yes, you probably have the beta1mod, not beta2.

[kenpachi]

I've returned to the clip for serious this time and got some other problems. How to get rid of vertical dirt VCR lines? I tried deVCR DeScratch() but with such a slight effect it's nothing like removing. Any tips?

example2: http://www.speedyshare.com/files/23435302/example2.rar

The script I used:

mpeg2source(".\example2.d2v")
checkmate(100)
greyscale()

Yadif(mode = 0, order = 1)
Crop(6, 0, -6, -8)

a = last
DeScratch(mindif=1, border=5)
DeSpot(p1=35, p2=2, pwidth=70, pheight=70, mthres=25, \
mwidth=20, mheight=15, interlaced=false, merode=50, \
ranked=false, p1percent=0, dilate=0, fitluma=false, \
blur=3, motpn=true, seg=0)
#DeSpot(p1=35, p2=14, pwidth=70, pheight=70, mthres=25, \
# mwidth=20, mheight=15, interlaced=false, merode=33, \
# ranked=false, p1percent=0, dilate=0, fitluma=false, \
# blur=4, motpn=false, seg=0, tsmooth=3)
#RemoveDirt(grey=true)

Undot()
GradFunkMirror(strength = 1.3)
#fft3dfilter(sigma=0.5, bt=3, dehalo=1)
#smoothlevels(35,1,255,0,255)
#grainfactory3(3,4,2,45,45,45,1.05,1.05,1.05, temp_avg=25)

#Tweak(sat=1.3,bright=-14,cont=1.25)
#ColorYUV(gain_y=110,off_y=-23,cont_u=76.8,cont_v=76.8)

StackHorizontal(a.Lanczos4Resize(704,464), last.Lanczos4Resize(704,464))

#Info()

[poisondeathray]

I've returned to the clip for serious this time and got some other problems. How to get rid of vertical dirt VCR lines? I tried deVCR DeScratch() but with such a slight effect it's nothing like removing. Any tips?


are you talking about the fairly constant 2 vertical lines ?

with descratch, the most important parameter for detection IMO is the minlen . Set minlen low e.g. minlen=5 . You can fine tune the other parameters , have look at the documentation. Don't use a border setting that high, or you will get weird artifacts. I would keep it lower or default (=2)

if it destroys other parts of the image too much, one way to limit the damage to those 2 lines (in addition to adjusting the other parameters) is to do an overlay or mask . e.g do a strong descratch setting video, and crop & overlay that on top of your normal denoised/despotted video

[kenpachi]

@poisondeathray: Thx for reply. That's it: minlen=5 does the trick, You're right.

I tried now:
DeScratch(mindif=1, minlen=5, border=1)

but even border=1 cannot preserve artifacts at some other moments which is unacceptable. You say the only way to deal with it is to combine Crop, overlay, trim... I get it but I'm too lazy to work on this clip anymore. DeSpot is enough.