Deinterlacer for analog capture

[yup]

Hi folks!
http://www.sendspace.com/file/0dfhlv
The source is analog capture from VHS tape. I try find deinterlacer:
first
AVISource("sel.avi")
AssumeTFF()
Telecide(guide=2)

Second
AVISource("sel.avi")
AssumeTFF()
tfm(slow=2, mChroma=false)

but both do not work fine. May be need preprocessing source file? Please advice.
With kind regards yup.

[ajk]

Hi,

Try

AVISource("...")
Telecide(guide=2)

or

AVISource("...")
AssumeBFF()

SeparateFields()
o=SelectOdd().Trim(1,0)
e=SelectEven()

Interleave(o,e)
Weave()

[Guest]

first
AVISource("sel.avi")
AssumeTFF()
Telecide(guide=2)
What is "not fine" about it? You have field-shifted PAL, so this is a correct script for that, and it appears to work well on your sample. So what is the problem?

You can use post=0 to turn off postprocessing if it is not needed.

[ajk]

There is a bit of difference in the results depending on whether one assumes bff or tff before telecide(), or does it manually, but it doesn't really make it or break it in my eyes between any of the three.

[Guest]

The sample was TFF and it makes a big difference to Telecide().

[ajk]

Not to hijack the thread, but as this now came up, what does telecide() do differently in this case to the manual field shifting? At least with this sample some frames end up not quite the same with the two methods.

[Guest]

Telecide() matches the fields using a heuristic. That means it can adapt to breaks in the pattern. For example, if the stream has sections that are and are not field-shifted, your method would require manual intervention.

[yup]

Hi ajk and neuron2!
I know that source BFF, problem with my capture card Aver TV 507, I capturing using iuvcr and huffyuv codec. Neuron2 problem notches on foot Samson and luma borders.
With kind regards yup.

[ajk]

@neuron2

Yes I realise this, but what I meant was that doing a subtract() out of the results of telecide() and the manual field shifting in this particular case does not result in a completely empty image with all of the frames, and I was just wondering where the small differences come from.


@yup

I think the deinterlacing works fine here (after all, the fields just need to be arranged to the correct order, nothing else), but you may need to do some other processing to clean out the image if you don't feel like the results are yet good enough.

If by luma borders you mean the halo effect, you could have a look at the many discussions regarding de-haloing, starting with this thread (http://forum.doom9.org/showthread.php?t=102794) for example.

[yup]

ajk!
Very interesting thing, your simple script:
AVISource("...")
AssumeBFF()

SeparateFields()
o=SelectOdd().Trim(1,0)
e=SelectEven()

Interleave(o,e)
Weave()
give better result than Decomb plugin, why?
yup.

[ajk]

I can't explain it, I would have imagined it would be exactly the same as telecide(). I don't think really think there is any significant difference, but if you feel the results are better that way, go with it :)

[yup]

ajk!
You can see difference using denoising, I use MVDegrain2. May be problem with source realy TFF have flag BFF.
yup.

[Chainmax]

I can't look at the clip right now, but take a look at the links in my signature. They point to threads in this forum and the developer of one of the most famous comb removal filters. It will take you a bit of time to get through all of it, but it's not overwhelming. If you have any doubts, don't hesitate to ask.

[yup]

Hi Chainmax!
May be I find reasin not very good work for telecide, interference noise.
AVISource("sel.avi")
AssumeTFF()
Crop(16,16,-16,-16)
SeparateFields()
DeFreq(info=true, show=1, plane=0)

see image
http://img150.imageshack.us/img150/4107/defreqzk2.th.png (http://img150.imageshack.us/my.php?image=defreqzk2.png)
I reead thread about Defreq, now this plugin not devoloping advice other which remove diagonal interference noise.
yup.