Benefit with DVCPRO50 and Vegas 6?

[Pesie]

Hi everybody,

I was thinking of an interview with someone who worked on the movie November, that was shot with a DVX100. I remember he/she told that they edited November on some high-end Avid workstation in uncompressed mode, to have better color correction.

When you scale this down to Vegas, is it possible and does it even has any benefit of using the Matrox DVCPRO50 codec with Vegas for my SD DVX100 footage. I have installed the codec and it's possible to render my captured DV25 AVI files to DVCPRO50 AVI files with Vegas. When I edit and color correct my film projects using these DVCPRO50 AVI files, is there any benefit from this? Like a more accurate color correction etc? When you render to MPEG2, the color depth that is available is (with PAL) 4:2:0, not 4:2:2.

Please shine any light on this matter, thanks!

[Pesie]

Oops, I discovered that you can use 4:2:2 rendering to MPEG2, when changing to the High Profile option in the Video section of the rendering settings. But has this any visible benefit?

[glenn chan]

In Vegas:
A- You could add the chroma blur, which will blur your color into 4:4:4 (sort of). It doesn't give you the best results possible, but it's reasonably good. To get the best chroma reconstruction, you want to look at techniques like:
--Summing the luma around where the chroma values are sited, to yield a luma+chroma pair. From there, you want to smooth out the chroma but make sure it's proportional that you account for differences in luma.
You can sort of see what I'm talking about in the following article:
http://www.glennchan.info/articles/technical/chromata/chromata.html
However, there is a major error in there (which I realized just recently) in which I didn't realize that this technique is better than what I was doing. It stems from my misinterpretation of how chroma sampling works... the chroma samples do/should not describe the chroma at a particular pixel, it should describe the chroma over an *area* of pixels. Vegas' chroma blur, in a way, has this misinterpretation built into it.

--More intelligent chroma reconstruction, that makes the assumption that areas of similar luma have similar chroma (which is generally true). Make this assumption should make the guessing of missing chroma values much more accurate.

B- The problem with chroma blur is that it's painfully slow. A slightly faster approach would be to use a DV codec that applies chroma interpolation. I don't know which ones these are off the top of my head.
The thing you do need to watch out for is that some codecs do computerRGB color space, which will result in a loss of information (studioRGB has less loss).

C- Another option is to batch render your edit into another format, with the chroma blur already applied. SonyYUV is 4:2:2, which won't give you the best results if you want to do any secondary color correction. A lossless format would be appropriate... maybe Huffyuv works.

Avoid lossy compression formats like DVCPRO50.

[Pesie]

Wow Glenn, this is like science fiction for me. I was thinking of a way to get a better way to handle color correction within Vegas. But reading your info, this is not an easy matter, right?

[glenn chan]

Yeah I kind of rambled on there...

The easy answer is:

Apply the chroma blur filter. For NTSC DV, use 2.0 pixels horizontal. For PAL DV (excpet DVCPRO), use 1 horiz. 1. vert.

Somehow you want to pre-render everything so that performance is good. I don't mean the actual pre-render.... I mean something like "render as..." and bring things back in as takes or whatever.
Save as and trim media might actually work too, but I'm not sure. There might be some easy scription solution.
But basically, you want something like:


- Render out your whole project to a lossless format (i.e. huffyuv) with chroma blur applied. Bring it back in, put the cuts in the right places. Handle dissolves specially.

Or... just apply chroma blur when it matters. Usually it doesn't. It matters a little for secondary color correction, and matters for chroma key.
So if you want to be lazy, just do everything without chroma blur. At the end, apply the chroma blur to your video output FX chain, and make your DVD (or whatever your deliverable is).

[Pesie]

Thanks Glenn, this isn't science fiction. I will do some testing.
Greetings,

[stokestack]

The answer to your question is no, unless you go to something with greater than 8 bits, which DV50 and MPEG-2 aren't. Not to mention the fact that going to either one would mean a unnecessary round of compression. Do not do a pointless render into a lossy format before you start color-correcting. That would gain you nothing and be quite destructive to image quality. To have any chance of gaining anything, you'd need to render your original DV25 out to something like the Sheer or Microcosm codecs and do your color tweaking in a 16-bit domain.

The 4:2:2 sampling of DV50 is only a spatial (color accuracy across pixels) advantage, which you've already lost by starting with DV25; it is not a more precise sampling of color values. It's not analogous to something like 24-bit vs. 16-bit audio sampling.

Your best bet is to do your color correction in your DV25 project and then render your final version into your delivery format (MPEG-2?) and an uncompressed or lossless archival format and be done with it.

[glenn chan]

The benefit in this case comes from "uprezzing" the chroma. DV is 4:1:1, where the chroma is a quarter the resolution horizontally compared to the luma. The Sony Vegas DV codec does not scale up the chroma nicely, leading to chunky color artifacts. It's like the "nearest neighbour" rescaling mode in Photoshop versus its other rescaling modes (that look better).

So you have a minor advantage in bumping up the chroma, and using an intermediary codec to get better performance (rendering it).

It's not a big difference, and mostly comes down to workflow.