Capturing YUY2 or YV12 interlaced?

[rmdeboer82]

Hello,

I want to convert some old VHS tapes to DVD with my old Medion MD2819 (sa7134 chipset) using VirtualVCR. I read several threads about it.

Should I capture with YUY2 (4:2:2) or YV12 (4:2:0)? The last one would be easier since HC encoder requires YV12 as input (using both interlace settings under 'chroma downsampling' and 'interlacing options'). Then I should use Lagarith encoder since it supports YV12. However, I will lose some information for the filters (due to 4:2:0). Besides that, does Lagarith has fully support for interlaced source (for both YV12 and YUY2)? I read that it can give some artifacts using YV12 interlaced but I don't if it has been improved.

Or is it better capture YUY2 with Huffyuv and use the ConvertToYV12(interlaced=true) AviSynth command for the input file for VDubMod and HC encoder? Then I got more info for the filters (4:2:2) and I did not read that interlaced is problem for Huffyuv. My aim is to get quality.

If I use the ConvertToYV12() command what does happen behind the scenes then? Will it request the encoder (Huffyuv or Lagarith for example) to translate it or does AviSnyth this internally? And when displaying it in VDub it gets translated to RGB for viewing purposes only (the raw material remains the same or is converted to YV12)? I tried in VDub to use YUY2 as input (via AviSynth) and YV12 as output (or the other way around) but it let not do me this (gives an error when the color spaces mismatches). So for output I should stick to the ConvertTo<XXX>() command that I gave in the avs input file, that's no problem.

I will use an intermediate file to prevent 'frames found with small mismatch errors' of the HC encoder (due to DirectShowSource in stead of AVISource avisynth command). So input and output for this intermediate file will be either YV12 or YUY2. However, the final file will be YV12 as required for DVD.

AviSynth's interlaced=true means that both input and output is expected to be interlaced, right? I read the site, but it's a very short description.

I read that VDub is updated, but that VDubMOD has support for the interlaced YV12. Is that still right?

Does anyone know if YUY2 or YV12 is the native format of this card (sa7134 chipset)? Because I read that the drivers can convert it on the fly to another color space with loosing detail/quality.

Related posts:
http://forum.doom9.org/showthread.php?t=128197
http://forum.doom9.org/archive/index.php/t-139828.html
http://forum.doom9.org/showthread.php?t=91451
http://forum.doom9.org/archive/index.php/t-84397.html
http://forum.doom9.org/archive/index.php/t-74303.html
http://originaltrilogy.com/forum/topic.cfm/What-codec-do-you-use-for-editing/topic/11652/

I will use QuEnc for separate encoding of audio.

Thank you for your help.

[jagabo]

All capture cards initially capture something like YUY2 (YUV with 4:2:2 chroma) because that's the closest thing to what's in an analog signal (a luma signal with two chroma signals at half or lower resolution). The drivers for some may then convert to YV12 or RGB if requested.

If you are going to perform extensive filtering you should capture as YUY2 and convert to interlaced YV12 just before encoding. That will preserve a little more color resolution.

[Brad]

If I use the ConvertToYV12() command what does happen behind the scenes then? Will it request the encoder (Huffyuv or Lagarith for example) to translate it or does AviSnyth this internally?

It uses its own internal filter.

To be honest I don't understand most of your other questions or don't know the answer but I will say that interlaced YV12 is a big bag of worms on its own and there is no way for any of the filters or programs to know whether they should be dealing with the chroma as progressive or interlaced unless there are explicit options for you to set. I would save yourself the headache and capture in YUY2 then convert to YV12 interlaced as the final step.

Or if they are tapes you can perform IVTC on then you can encode progressive.

[jagabo]

AviSynth's interlaced=true means that both input and output is expected to be interlaced, right?

Yes.

I read that VDub is updated, but that VDubMOD has support for the interlaced YV12. Is that still right?

If you open an interlaced MPEG file directly with VirtualDubMod it will convert to RGB with proper handling of the interlaced chroma channels. VirtualDub will not. Both will screw up if you feed them interlaced YV12 from AviSynth. I always use AviSynth to open any interlaced YV12 source and ConvertToYUY2(interlaced=true). If you use DgIndex you can have it to do the conversion for you with Mpeg2Source("file.d2v", upconv=1). For example:

Mpeg2source("file.d2v")
ConvertToYUY2(interlaced=true)

or

Mpeg2source("file.d2v", upconv=1)

[2Bdecided]

does Lagarith has fully support for interlaced source (for both YV12 and YUY2)? I read that it can give some artifacts using YV12 interlaced but I don't if it has been improved.

Lagarith is a lossless codec. It doesn't change anything. I've never heard any reports to the contrary. I always use it with YV12.

[2Bdecided]

If I use the ConvertToYV12() command what does happen behind the scenes then? Will it request the encoder (Huffyuv or Lagarith for example) to translate it or does AviSnyth this internally?

It uses its own internal filter.

To be honest I don't understand most of your other questions or don't know the answer but I will say that interlaced YV12 is a big bag of worms on its own and there is no way for any of the filters or programs to know whether they should be dealing with the chroma as progressive or interlaced unless there are explicit options for you to set. I would save yourself the headache and capture in YUY2 then convert to YV12 interlaced as the final step.

...but plenty of useful AVIsynth scripts and filters only work in YV12 colour space.

AND most scripts or filters will only work properly with interlaced or progressive - there aren't many "one size fits all" scripts - so whether it's YUY2 or YV12, you need to use scripting that works with interlaced footage if that's what you have.

It goes without saying (but I'll say it anyway) that scripts which (a) only work in YV12 colour space, and (b) are written to work with interlaced footage, are going to handle interlaced content properly in YV12 colour space.


If DVDs were 4:2:2 then using 4:2:0 would be an abomination that should be avoided. But since we're stuck with 4:2:0 for all output then there's no harm using it throughout (unless you do something stupid!!!).

The one possible benefit from 4:2:2 is to keep maximum chroma resolution if you intend to upscale - but many of the useful AVIsynth functions relating to upscaling are 4:2:0 YV12 only, so in practice there's still no point. Not unless you code around this limitation, and/or want to capture a still image (which may be a 4:4:4 RGB TIFF).


So, long story short: 4:2:2 really should be better in theory, but I can't find a (non-broken) workflow where it makes any difference. YMMV of course.

Cheers,
David.