Colour Space. Which?

[barryland]

I am currently working my way through my collection of shockingly crap martial arts films and I need to use various filters to de-interlace, resize, logo and sexify my ugly-ass video.

I just can't figure out what colour-space I should be capturing in. I use virtualdub for the cap and the format conversion and but have noticed all sorts of nasty artifacts when converting my Mjpeg captures to Xvid using Vdub's "full processing" option.

Can anyone tell me what the best colour-space to capture in would be? Or even in fact if I am barking up the wrong tree all-together.

(Sorry if this is a bit confused but I've been wading through Google translated web pages looking for an answer and my brain is fried)

[trevlac]

I can't say I'm an expert in this, but I have been reading a bit lately. Here is what I think I know.

The original source is basically YUV which in codec terms is YUY2 or UYVY or YVYU which are all 4:2:2 subsampled (or basically the same as the source).

So if you look here (http://www.virtualdub.org/docs_capture) you'll see a recomendation to cap in one of the YUV flavors.

However, I believe Xvid or any MPEG compression uses 4:2:0 subsampling. (Mpeg2 does. You'd have to check on mpeg4 which I don't do.) Your card can probably do a decent job converting down to this which it will do as YV12. This is also the YUV color space so it is just a matter of removing some of the info on every other line.

Finally, It is my understanding that VirtualDub in full processing operates in the RGB color space. (As opposed to the YUV one.) I take it this means it converts from YUV to RGB whenever you use a filter. So if you use VDub full processing, from an mjpeg cap of YUV to an Xvid of YUV you are converting to RGB in the middle.

Disclaimer:
I don't do mpeg4 and I have not really worried about the colorspace thing when I filter my YUY2 caps. So maybe someone else can give you the 'right' thing to do. I just wanted to point you in the right direction. :)

Here (http://www.avisynth.org/index.php?page=DataStorageInAviSynth) is a link I found informative recently. As I post it, it does not work, but I'm sure it will be back.


Below is a forum topic that tells me someone caps in YV12, uses Avisynth to filter, and uses virtualDubMod fast recompress to encode YV12 to Xvid. No color space conversion. Just a downsample (by the card or maybe avisynth).

http://forum.doom9.org/showthread.php?s=&threadid=57608&highlight=rgb24