One more question about halo removing

[yup]

Hi folks!
My source is analog capoture from retail VHS. I see oversharpening at luma changes. This is halo?
See one frame
http://thumbnails2.imagebam.com/2718/2bb83027177727.gif (http://www.imagebam.com/image/2bb83027177727)
and 100 frames
http://www.sendspace.com/file/uiicir
I try DeHalo_alpha, BlindDeHalo3 and FFT3Dfilter with dehalo parameter and do not satisfy. Please advice.
With kind regards yup.

[MadRat]

It's hard to be sure from the still image but yes there are halos in your video. Carefully watch the space just outlining the soldier's head and you'll see a halo.

[yup]

Hi folks!
After googling I understand what deal with overshooting and undershooting video signal after low pass filtering.
Question:
can I try find parameter low pass filter and repair video?
With kind regards yup.

[yup]

I try also YATHR:
http://forum.doom9.org/showthread.php?p=1205653#post1205653
but without success. May be main problem very long distance halo and this case require other approach.
yup.

[leeperry]

it looks like VHS ghosting more than EE halo's, I think it's beyond repairability :o

[yup]

it looks like VHS ghosting more than EE halo's, I think it's beyond repairability :o
What need make for remove this ghost?
yup.

[leeperry]

there's a VDUB plugin here, might wanna give it a shot maybe:
http://www.hlinke.de/tinc?key=9An613gk&session_currentpage=data&session_mode=guest&formname=VirtualDub_Filter&session_sortby=field_2&userid=1204388450;283847;2&session_nextpage=data_edit&session_offset=20&session_start=21&session_dbkey=1154804372;105345;89_VirtualDub_Filter&dbkey=1154804372;105345;89_VirtualDub_Filter

[Didée]

Hey, removing those halos is pretty easy!

http://thumbnails5.imagebam.com/2870/4f09fd28691616.gif (http://www.imagebam.com/image/4f09fd28691616)

The hard part is to make anyone like that result ... ;)


gaussresize(512,576,p=8).gaussresize(360,576,p=6).gaussresize(360,560,p=12).spline36resize(720,576)

[yup]

leeperry & Didée :thanks:
I will be test.
yup.

[Mug Funky]

@ Didée:

maybe sit very far back from the TV?

i think with gibb's style ringing that low in frequency there's not much chance of resolving any real detail beyond what you've got. the best course from here is to choose the form of aliasing or distortion that looks best...

[yup]

@ Mug Funky!

I agree problem with low pass filtering during VHS recording. Question how this recover. May be try apply high pass filter which will be near inverse loaw pass filter and after antialiasing filter.
yup.

[Didée]

Oh well. Switching back to the unloved (to me) hull-edgemask way:

http://thumbnails5.imagebam.com/2879/34299228784218.gif (http://www.imagebam.com/image/34299228784218)

Quite reasonably on wide parts, but (necessarily) inconsistent in places where fine detail is located next to high-contrast edges (look in the lower left, the green stuff in the background, between the wooden poles in the foreground).

Needs a bit more tweaking, but generally this is about as good as it'll get. Script to come, after cleaning.

[*.mp4 guy]

mt_convolution(horizontal=" -4 -8 44 -20 -12 384 -12 -20 44 8 -4 ", vertical=" 1 ", u=2, v=2)

YAHR()

MT_Merge(mt_convolution(horizontal=" 0 -12 -28 128 -28 -12 0 ", vertical=" 1 ", u=2, v=2).mergechroma(last, 1), last.mt_convolution(horizontal=" 0 0 4 40 -2 -8 128 -8 -2 40 8 -4 2 ", vertical=" 1 ", u=2, v=2).YAHR().mergechroma(last, 1), last.mt_convolution(horizontal=" -4 -8 44 -20 -12 128 -12 -20 44 8 -4 ", vertical=" 1 ", u=2, v=2).halomask1(ss=1.75, rad=8).frfun7(1.49, 256, 0))

or, less haloing, more aliassing, more blurring, more correct frequency distribution:
mt_convolution(horizontal=" -4 -8 44 -20 -12 160 -12 -20 44 8 -4 ", vertical=" 1 ", u=2, v=2)

YAHR()

MT_Merge(mt_convolution(horizontal=" 0 -18 -12 128 -18 -12 0 ", vertical=" 1 ", u=2, v=2).mergechroma(last, 1), last.mt_convolution(horizontal=" 0 0 4 40 -2 -8 128 -8 -2 40 8 -4 2 ", vertical=" 1 ", u=2, v=2).YAHR().mergechroma(last, 1), last.mt_convolution(horizontal=" -4 -8 44 -20 -12 160 -12 -20 44 8 -4 ", vertical=" 1 ", u=2, v=2).halomask1(ss=1.75, rad=8).frfun7(1.49, 256, 0))

[yup]

Didée & *.mp4 guy thanks!
will be testing.
at *.mp4 I will be try understanding how work FIR (finite impulse responce ) filter.
yup.

[Didée]

Oh, yes ... that's what happens when you want to clean a script, but instead you come into toying and make the script longer instead of shorter ...

normal - http://thumbnails16.imagebam.com/2922/0e078529215484.gif (http://www.imagebam.com/image/0e078529215484) . . . . . sharper - http://thumbnails7.imagebam.com/2922/0a239129215485.gif (http://www.imagebam.com/image/0a239129215485)

I'll try to cut it down, finally. ;)


(edit) - BTW, with mp4guy's scripts it seems I get the following results:

first script - http://thumbnails6.imagebam.com/2922/c9b41f29216731.gif (http://www.imagebam.com/image/c9b41f29216731) . . . . . 2nd script - http://thumbnails2.imagebam.com/2922/ac1b5d29216732.gif (http://www.imagebam.com/image/ac1b5d29216732)

Is that the expected result?

[*.mp4 guy]

Yes those are the expected results. Even with all of the artifacts, I tend to prefer the last one I posted, though most people probably wouldn't.

[edit] TBH, I'm not entirely happy with the results (nor Didée's, though they are much more consistent). The halo's are just too strong for truely good results with any close to prepackaged method I can think of.

[yup]

Didée & *.mp4 guy :thanks:
Didée :script: for halogone3.png. I try encoding small file and looking at TV.
yup.