Home movies Deinterlacing

[drob]

This is a split i made form my original thread at the x264 forum
http://forum.doom9.org/showthread.php?t=152804

The short version is that my dad captures home movies using an mpeg2 camera, to archive them i am using StaxRip, i was advised to keep the footage interlaced, but since Pinnacle Studio is having hard time reading the resulting interlaced MP4 file, i need to Deinterlace before hand.

StaxRip has these filters for Deinterlacing:
yadif
FieldDeinterlace
TomsMoComp
SelectEven
SeparateFields

Considering this is meant for archival purpose, what would you suggest? any other better filter i should use? i tried yadif with mode 1 and it looked nice, but not as sharp as the original.

[davidhorman]

If it is for archiving, you should just keep the MPEG2 files - though I'm assuming you don't have the space, which is why you're converting. Perhaps you could compromise and use, say, Xvid, which Pinnacle might have better luck with?

David

[Didée]

Short of reading the other thread (am lazy now), somehow the short scenario makes little sense. If editing is to be done, the normal workflow would be mpeg2->editing->recompress(archival). This workflow does not require Pinnacle to read mp4. Your short scenario sounds more like mpeg2->recompress(archival)->editing. Which is awkward. Did I miss something important from the other thread?

[drob]

My bad for not being clear enough, the workflow is: Mpeg2-editing-(and the killer) until now my dad saved it as uncompressed avi. thats where i stepped in and thought it would be much much better for future reference to keep the archival copy as H264.
The reason for needing Studio to read the files is for extra copies and so on.

Now since Studio seems to have trouble reading the pure interlaced encoded H264 in the MP4 container, i need to deinterlace and keep it all progressive. For my dads sake i had to find a nice simple GUI which i can pre program to do all the chores, thats where StaxRip came in.
So the question remains which of the above filters would serve best for the purpose, or which other filter would you recommend, and with that settings.

[nm]

The reason for needing Studio to read the files is for extra copies and so on.
What are these extra copies exactly? Can't you use some other tool to do this?

So the question remains which of the above filters would serve best for the purpose, or which other filter would you recommend, and with that settings.
I wouldn't use the filters offered by Staxrip. If you have an Nvidia 8xxx or newer GPU with VP2/3/4, DGDecNV (http://neuron2.net/dgdecnv/dgdecnv.html) would be a good and fast choice, but its deinterlacing quality may depend on driver settings. There's seems to be some discussion about using it with StaxRip here: http://forum.doom9.org/showthread.php?t=151118

As for AviSynth filters, perhaps (noise-bypassed (http://forum.doom9.org/showthread.php?p=1324744#post1324744)) TempGaussMC would be appropriate?

[hydra3333]

Maybe see http://forum.doom9.org/showthread.php?p=1374335#post1374335 and the few posts after that. Don't forget to view Didee's sample found there, it might change your mind on the deinterlacer.

[2Bdecided]

If using an MPEG-2 camera, my first preference would be to use an NLE that supports smart rendering with SD MPEG-2, so on the portions of the video that have been untouched (and unless your Dad is doing colour correction on the entire project, that'll probably be most of it), the output is just a straight copy of the input - compact, fast to render, and lossless. The little sections around edits, and with titles, will get re-encoded to similar MPEG-2.

That output MPEG-2 will be fine for DVD (though 8.5Mbps DVD-R may be unreliable in some players), and fine for archiving.

Job done.


If you insist on using a work flow without smart rendering, then you're stuck with either:
1. lossless (HuffYUV or Lagarith are better than uncompressed lossless)
2. some proprietary near-lossless codec (good choice for quality vs size vs future re-use, but not free, and might confuse anyone using the files in five years time without the right codec)
3. DV (free, fast, ubiquitous, decent quality, but may compromise colours on NTSC footage slightly)
4. High bitrate MPEG-2 (slow encoding, OK quality)
5. High bitrate MPEG-4 (slow encoding, interlacing support is dubious - re-editing may be problematic)
6. High bitrate MPEG-1 (lousy interlaced support)

Personally, I'd go with HuffYUV or DV.


If you decide to deinterlace, but may later want to re-interlace, it makes sense to use a deinterlacer that's lossless on the lines that will get used for the re-interlacing. The deinterlacer that I'd normally suggest (TGMC) isn't lossless in this way.

Cheers,
David.

[Blue_MiSfit]

I'd suggest using YADIFMod + NNEDI, or something like this. TGMC will probably give better quality, but as 2bdecided mentioned, it may be problematic for a case like this (doubtful though)

[hydra3333]

yadifmod I'd thought OK until I saw this...
Partly because the "faster" TGMC settings hold up so well with that sample (going to try them on some of mine now!) - but even more because it shows that yadif isn't magic at all. I keep hearing people describe it as "pretty good" - it's not - if anything moves, it's just "bob"! Yadifmod means it's just NNEDI2 in this case, but it fails so badly on this sample. IIRC EEDI2 is happier connecting the lines on that one, but still pails in comparison with TGMC.

[drob]

I agree that TGMC does look better then yadif, however luckily i managed to solve my issue with opening the interlaced encode in Studio (using m2ts instead works), so i do not need the deinterlacer no more.

What i would like is to add some denoising, i liked the VagueDenoiser included in StaxRip, any pointers on settings for some general light denoising , attached is a sample, ignore the combing.

http://img714.imageshack.us/img714/1170/vlcsnap2010022221h59m47.th.png (http://img714.imageshack.us/i/vlcsnap2010022221h59m47.png/)

[2Bdecided]

I'd suggest using YADIFMod + NNEDI, or something like this. TGMC will probably give better quality, but as 2bdecided mentioned, it may be problematic for a case like this (doubtful though)Yeah, I'd use it anyway - the implicit denoising might help. The OP has moved on though!

[2Bdecided]

What i would like is to add some denoising, i liked the VagueDenoiser included in StaxRip, any pointers on settings for some general light denoising , attached is a sample, ignore the combing.

http://img714.imageshack.us/img714/1170/vlcsnap2010022221h59m47.th.png (http://img714.imageshack.us/i/vlcsnap2010022221h59m47.png/)That's not a sample, that's an image! ;)

Anyway, Google is your friend - first hit for VagueDenoiser is the AVIsynth version...
http://avisynth.org.ru/vague/vaguedenoiser.html

Never used it. Looks interesting and fast.

Most people seem to use variants of mvdegrain or fft3dfilt. There are lots of older noise filters for AVIsynth which are sometimes useful, e.g. deen.

http://avisynth.org/mediawiki/Denoisers

Cheers,
David.

[drob]

According to its webpage, mvdegrain doesn't support interlaced material, so no good for me:

Only progressive YV12, YUY2 video is supported. Use color format conversion and try use (smart) bob-deinterlace for interlaced video

Ill give FFT3DFilter a try.

As for VagueDenoiser i did use Google (i am not that lazy...) just wanted some recommendations.

[drob]

FFT3DFilter is really nice not only it cleans the image but details remain nice, and it sharpens a bit, so thats great, thank you!
Took e a while to find how to integrate it into StaxRip, but once you get the hang of it its easy.

[2Bdecided]

According to its webpage, mvdegrain doesn't support interlaced material, so no good for meYou quoted the bit that told you exactly how to use it with interlaced material!!?!

Still, glad you're happy with fft3dfilter.

Cheers,
David.

[davidhorman]

A nice way to use mvdegrain or mdegrain with interlaced material, if you want it to remain interlaced afterward, is to separate the fields, then split the top and bottom separated fields into two clips, run each through mdegrain, then unsplit and reinterlace the cleaned clips. This way no "invented" information is used during processing.

David

[Didée]

separate the fields, then split the top and bottom separated fields into two clips, run each through mdegrain, then unsplit and reinterlace the cleaned clips
Correct, but a lil complicated. Easier: Separatefields, then do MSuper/MAnalyse(..,delta=2,..)/Mdegrain, then weave.

[davidhorman]

Well there you go, now I've learned something too!

David

[2Bdecided]

On my S-VHS camcorder footage, that separatefields trick doesn't work. You can't see it in the original (noisy) footage, but the noise and other artefacts in the footage are somewhat (but consistently!) different in the two fields. If you run the denoiser on separated fields, and then weave them back together, you get combing or flickering (depending on the display) quite visible in the clean footage. Whereas if you denoise bobbed footage, it's fine - the denoiser works across both fields, and creates a denoised version that's consistent between fields.

This may be a near-unique situation, but it's worth being wary.

Cheers,
David.