Selur's Little Message Board

Full Version: De-Halo help
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
I've tried both de-halo and de-ring filters but I have not been able to remove these halos, which are on a PAL broadcast TV master from 1992.
It appears to the right of his dark jacket when against a lighter background.

Any suggestions?

[Image: Halo-sample3.png]
At an Iron Maiden concert atm. will look at it tomorrow.
Cu Selur
Beast World!
Yup, 'Legacy of the Beast World Tour'
Taking the image you posted as source, DeHalo_alpha seems to work:
[Image: dehalo.png]
it seems like it's just a question of finding the right values.
  • enable 'Filtering->Synth auto refresh',
  • enable 'Filtering->Filter view',
  • 'Filtering->Vapoursynth->Misc->Split view' ,
  • set 'Filtering->Vapoursynth->Misc->Split view' to 'interleaved', and
  • set set the a fixed zoom in the previewer i.e. 4
(dehalo strength is rather sensitive, you usually want to keep it below 2)
this way it you can better see the effect of the dehalo filter, combined with some additional denoising&co this probably works out fine.

Cu Selur
Hmm, it looks like the main halo is still present, along the right side of the jacket sleeve.
I didn't try to tweak the settings, I just tried the first dehalo filter saw that it worked and suggested using it together with some denoising,... (I used the image you posted as source, so can't really use temporal filtering Wink)
Since I don't have a video sample, I used the image and know stumbled over a problem when applying masking on RGB content. -> working on that
I tried this dehalo filter with numerous different settings, but I couldn't get it to work.
Would need a small sample clip with the issue, to have a temporal component to play with and suggest anything meaningful.

Cu Selur
Probably not really helping if apply ctmf to a specific region, frame range could be added, you could get rid of the halo,....:
# cropping the video to 736x572
clip = core.std.CropRel(clip=clip, left=8, right=24, top=2, bottom=2)
# denoising using CTMF
def vsCTMFRegion1(clip,*args):
  return core.ctmf.CTMF(clip=clip, radius=3, memsize=1048576)
end = clip.num_frames-1
mapping = [
  (0, end),   [vsCTMFRegion1],
clip =, mapping, selection=(156,288,580,260))
[Image: ctmf.png]

Cu Selur