shaky video (not footage) with drop outs

[mammo1789]

I have one tape that has drop outs but more annoying is shakiness that is happening and the scene is quiet and static in studio.
I tried devcr for drop outs and it sorted some how, but the shakiness is there, there is also some other problems with it.
It is the same on other vcr's and the best pic is this from the panasonic nv fs200 "raw" lagarith footage unprocessed

[themaster1]

Try this script i've made (still based on poisondeathray's script), you're gonna like it i think:
script:
http://www.mediafire.com/?oegois9ogszmljm
video:
http://www.mediafire.com/?1xlvclds5zfp91f

This time i use Tmedian2 (i'm not a fan of it because when there is motion is creates morphing artefacts (maybe because of mvtools settings) but here it's a static scene, no motion or very little it work wonderfully imo
I also have added degrainmedian (it help to remove the white lines and ccd (help for the chroma noise)

If poisondeathray comes around here maybe he'll enlighten me on Tmedian2 and i'd like to know more about selectevery(5,2) that's for a pal/tff source i believe, what about ntsc or film sources ... i'm such a noob in this field

[poisondeathray]

LOL ..... again it's not "my script" . I mentioned this in the other thread - I got it from pirej who got it from Didee. (Almost all novel scripts are related one way or another to Didee) . So you should ask jedi-master-extreme Didée

Here is the 1st appearance of the script - yes it is damaging. I would ask the grand master himself for help . I'm not even a "padawan " . My brain hurts trying to decipher 1/10 of the things he does

http://forum.doom9.org/showthread.php?t=159914
post #10

[themaster1]

Thanks for the info i'll get in touch with him

[mammo1789]

Thanks master1 i will try it when I'm done reformatting my pc's (3 of them this evening ) and re installing avisynth. I tried win 8 but didn't like it at all so back to win7

LOL ..... again it's not "my script" . I mentioned this in the other thread - I got it from pirej who got it from Didee. (Almost all novel scripts are related one way or another to Didee) . So you should ask jedi-master-extreme Didée

Here is the 1st appearance of the script - yes it is damaging. I would ask the grand master himself for help . I'm not even a "padawan " . My brain hurts trying to decipher 1/10 of the things he does

Yes i also get it from pirej last time i used it, but you remind as for that script on this forums that's why we say poisondeathray

Edit I tried it tonight and it did solve some drop outs ( although missed some which can manually be corrected nut it will be pain in the ass to do it for 1h:30 min video). I don't know why you used camera color denoise from virtual dub in the same pass as the drop out script i think it is better to do first thing the script than corrected video in vdub and doing another filters (NV and camera color noise ) then export that from vdub and do it in some editor like premiere or sony vegas for instance.

Also i noticed you use a lot od loadplugin calls that are unnecessary for this script

But one thing i asked is still there that is the "shakiness or jerkiness" from the video not from the camera shakiness and it seems depan didn't solve it. I wonder if stab can solve this short jerkiness that occurs in the picture

[mammo1789]

I used DeVcr 3 times with different settings then stab

AVISource("")
AssumeTFF()
Crop(10,0,-10,-12)
AddBorders(10,0,10,12)
ColorYUV(gain_y=7,gain_u=6,gain_v=-10,cont_y=10)
SeparateFields()
function DetectVCRLines(clip c,int threshold)
{
spacial_data = GeneralConvolution(ConvertToRGB(c),0,"0 -1 0 0 2 0 0 -1 0")
bar_data = ConvertToRGB(BilinearResize(spacial_data,16,c.heig ht))
st_data = Overlay(bar_data,Trim(bar_data,1,0),mode = "subtract")
st_data2 = Greyscale(Levels(st_data,threshold,10.0,threshold+ 1,0,255,coring = false))
st_data3 = Greyscale(Levels(st_data2,127,10.0,128,0,255,corin g = false))
st_data4 = Overlay(st_data3,st_data3,y = -1, mode = "add")
return PointResize(st_data4,c.width,c.height)
}

function deVCR(clip c,int threshold)
{
mybars = DetectVCRLines(c,threshold)
return Overlay(c,Trim(c,1,0), mask = mybars,greymask = true)
}
DeVCR(30)
DeVCR(10)
Devcr(20)
################################################## ############################
#Original script by g-force converted into a stand alone script by McCauley #
#latest version from December 10, 2008 #
################################################## ############################

function Stab (clip clp, int "range", int "dxmax", int "dymax") {

range = default(range, 1)
dxmax = default(dxmax, 4)
dymax = default(dymax, 4)

temp = clp.TemporalSoften(7,255,255,25,2)
inter = Interleave(temp.Repair(clp.TemporalSoften(1,255,25 5,25,2)),clp)
mdata = DePanEstimate(inter,range=range,trust=0,dxmax=dxma x,dymax=dymax)

DePan(inter,data=mdata,offset=-1)
SelectEvery(2,0) }
Stab()
LSFMod()
Weave()
#Histogram(mode="levels")
ConvertToRGB32(matrix="Rec601", interlaced=true)

Then used Neat video and camcorder color n and MSU smart sharp.

The resulted file was compressed with mainconcept reference into DVD compliant Mpeg2 2 pass 8500 VBR 9500 Max

[themaster1]

It's not bad but not better than what i did and the colors are way off. It would be worth to try your script without neat or msu smart sharp and rgb convertion to see.
If you zoom-in you can see a problem on the fields (more obvious on the logo part)

[sanlyn]

Thanks to poisondeathray for passing on the script and to theMaster1 for mods made to it. I'll understand it one of these days. Does a lot of fixing, even if frames 55-56 have a streak and there's some frame hopping. Every clip lately has been a learning experience.

This script and Didee's original version are giving me a VirtualDub error several times during the run:
"Internal error: SSE state was bad before entry to external code at", the "at" followed by references like:
"\source\w32\videocodecpack.cpp:158", "\source\AVIReaderHandlerTunnelW32.cpp:153", and a couple of others. The VirtualDub forum has old posts advising to ignore the error -- which I did, and the results look OK. Can't say I've ever seen that one before.

In any case, still playing with this one. The color and levels are a problem. And whatever happened to the detail in the guy's shirt collar? Looks like a levels problem somewhere along the line in earlier processing? Those fuzzy lines and edges in the original are giving me fits. Anyway, I'm still at it.

[sanlyn]

Wow--What a problem child this video is! I think the color/levels/dark blocks in the beard are the worst. Got it somewhat more stable, thanks to pdr and theMaster1. Pulled so many tricks I don't even remember what I did! ReplacedFramesMC helped a bit with 6 frames. Not real happy with it, but for what it's worth.....



http://dc206.4shared.com/download/GGGLBS1x/branco_ccd2.avi

That troublesome background. Unfortunately it'll come right back as blocks, banding and grain when it's encoded. Gotta learn to use the dither plugin.

The image above illustrates why contrast and colors will never look quite correct with this video. It is apparently duped tape, and was at some point played on a machine with inappropriate IRE settings. Luma is cut off in YUV and RGB (at precisely RGB 30) -- "cut off" meaning that there is no luma data below RGB 30 (the same goes for Red). Add an earlier dark luma crush indicated by the sharp RGB 30 peak in RGB below, and you can see why the shadows will always be clumps with no detail, no matter how light or how dark one makes them. How the brights were wiped out is anyone's guess, but they were blown away long ago with bright green being clipped first (so "whites" or skin highlights will always be off, but both will never look correct at the same time). So if three objectives are realistic contrast, convincing color, and denoising, forget about the first two.

source frame 112, unprocessed, black edges removed: