View Full Version : ChubbyRain (rainbow remover) breaks on fast moving scenes
bizz & buzz
28th September 2009, 23:30
I've been using the modified version of ChubbyRain (HERE (http://forum.doom9.org/showthread.php?p=589885#post589885)) in order to remove rainbows.
It's great but there is one problem:
On fast moving scenes it breaks. It seems that on some frames it kills the red parts way too much...
source:
http://img17.imageshack.us/img17/8968/26347918.jpg
source + ChubbyRain2
http://img17.imageshack.us/img17/7485/11542019.jpg
Is there any way to improve the filter in order to prevent this?
Here is a sample clip:SAMPLE (http://www.mediafire.com/download.php?zywywyzom3d)
Two points:
1) After applying IVTC the issue is much more visible.
2) In this specific sample there are no rainbows, but I can assure you there are plenty in the rest of the clip.
bizz & buzz
30th September 2009, 13:40
Does anyone have any idea what's going on with this? it's driving me nuts :confused:
Didée
30th September 2009, 13:56
Looking at the avs, ChubbyRain is using brute force filtering for rainbow reduction. Now, brute force filters tend to give bruteforced results ...
If a medicine has too much adverse effects, stop taking that medicine.
bizz & buzz
1st October 2009, 01:08
Well, other rainbow removers I've tried (I admit - only on their default settings), seemed to either change the clip colors more then CR, or reduce the rainbows less then CR... So I'm at lost here :confused:
I've already tried SmartSSIQ, DeRainbow, and LUTDeRainbow, but I'll try them again if you'll recommend it.
Here is a small sample with the rainbows. Check out their white shirts: HERE (http://www.mediafire.com/download.php?wbmiwd1uukm) (1.42MB)
poisondeathray
1st October 2009, 01:22
Here is a small sample with the rainbows. Check out their white shirts: HERE (http://www.mediafire.com/download.php?wbmiwd1uukm) (1.42MB)
This example isn't fast moving, and chubbyrain2() seems to work ok with it, and it didn't screw up the 1st sample from what I could tell.
EDIT: ok , I see it's barely visible. Another option then, is to apply filter in segments only e.g. using Trim()
original
http://i37.tinypic.com/315oqdf.png
chubbyrain2()
http://i38.tinypic.com/343rdqv.png
poisondeathray
1st October 2009, 01:30
This is the same frame as yours in the 1st post, using chubbyrain2, and there is not that much discoloration as yours ??
http://i33.tinypic.com/2rr5vl3.png
I don't think this is any different than the one in your link, except adapted for MT versions for "chubbyrain2"
function ChubbyRain2(clip c, int "th", int "radius", bool "show", int "sft")
{
#based on Mug Funky's ChubbyRain
th = default(th,10)
radius = default(radius,10)
show = default(show,false)
sft = default (sft, 10)
u = c.utoy()
v = c.vtoy()
uc = u.mt_convolution(horizontal="1",vertical="1 -2 1",Y=3,U=0,V=0)
vc = v.mt_convolution(horizontal="1",vertical="1 -2 1",Y=3,U=0,V=0)
cc = c.mt_convolution(horizontal="1",vertical="1 2 1",Y=2,U=3,V=3).bifrost(interlaced=false).cnr2().temporalsoften(radius,0,sft,2,2)
rainbow=mt_lutxy(uc,vc,Yexpr=string("x y + "+string(th)+" > 256 0 ?")).pointresize(c.width,c.height).mt_expand(y=3,u=-128,v=-128)#.blur(1.5)
overlay(c,cc,mask=rainbow)
show==true? rainbow : last
}
bizz & buzz
1st October 2009, 08:41
Poisondeathray, problem solved thanks to you!
After you used ChubbyRain with no color problems, I realized that the cause of the problem lays not on CR, but on it placement in the script. While I was using it right after "MPEG2Source" and before any IVTC (as recommended here (http://forum.doom9.org/showthread.php?p=589012#post589012)), you probably placed it after IVTC... and it indeed worked - no color errors at all :)
Only problem is that now CR seems to be less efficient in killing rainbows, no matter how much I play with it's parameters :(
So back to square one...
Are there any filters/methods you guys would advise to try?
MadRat
1st October 2009, 09:29
If you're working with NTSC, interlaced video you could use tcomb. You know it's got to be good if tritical made it. Chainmax says you can use FFT3DFilter on chroma planes. I've never tried it and I'm not sure how to do it, but he's pretty good so he's probably right.
Blue_MiSfit
2nd October 2009, 05:14
fft3dfilter(plane=3, sigma=4, sigma2=5, sigma3=5, sigma4=4)
Adjust the sigma values until you get something to your liking. Each corresponds to a different frequency band.
plane=0 == luma
plane=3 == chroma
plane=4 == luma + chroma (default)
~MiSfit
bizz & buzz
3rd October 2009, 13:29
MadRat and Blue_MiSfit, thanks for your advices. After testing TComb and FFT3DFilter, it seems that ChubbyRain2(sft=14),although far from perfect, still gives better results on this clip.
:thanks: again for your help guys.
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.