YUV and RGB vs Filters

technobaba wrote on 9/27/2006, 11:27 PM
Does Vegus use YUV or RGB inside? Filters?
DV is YUV, so does Vegas operate internally in RGB?

I know that the camera captures in RGB and then converts that to YUV on tape. When are the colors lost? I've read conflicting stuff on the web. One said RGB throws away 75% of the original YUV colors. One said the reverse. Which is true?

What workflow do you use to retain maximum quality vis-a-vis conversion and filtering.
Can someone point me to a summary? more info?

Comments

farss wrote on 9/28/2006, 12:17 AM
Vegas uses RGB internally.
No workflow is going to make much difference, pretty well all the damage got done during the DV sampling in the camera.
The term YUV is largely meaningless.
In the broadest sense A YUV system can represent every possible color. However THE YUV system(s) used in video cannot, for practical reasons it's limited, subsampled etc, etc.
And of course at the end of the day what video gets displayed on cannot show the full range of colors either.
In all a fairly complex topic with several facets to it, many long and protracted discussions on it on this forum, many useful links in them as well.

Put it this way, just to keep it simple.
Vegas does a more than adequate job with video from any video camera. Take a camera costing $100K, process the footage through any of todays NLEs, doubt you'll see any difference related to color.

Bob.
GlennChan wrote on 9/28/2006, 9:30 AM
Take a camera costing $100K, process the footage through any of todays NLEs, doubt you'll see any difference related to color.
Well there are extremely subtle differences, but you probably won't see it (except for the superwhites issue).

The biggest difference is how the NLEs handle superwhites... white values above legal range. Some NLEs have pipelines that clip at 100%.

In FCP, you have bizarre superblack problems in that the scopes clip the superblacks (although they may or may not be clipped). No one seems to notice this.

Beyond that, there are subtle differences in chroma subsampling when you render. Basically, the color is reduced in resolution to lower bandwidth/cost/data rates. It's like taking an image in Photoshop and making it smaller. However, when you process the color, you have to "make it bigger" (or resample it). Since there are different ways to make an image smaller/bigger (i.e. Photoshop has 4-5+ methods), there are some subtleties between different codecs.
codecs.onerivermedia.com has some great pictures showing the differences.

GlennChan wrote on 9/28/2006, 9:36 AM
Colors are lost when you have have to round values to an integer.

The formulas are:

Camera originates with 12-bit or 14-bit or similar analog digital converter. RGB domain, linear light.

Gamma correction is applied to the signal. See Rec. 601, sRGB, and Rec. 709 transfer functions.

Then R'G'B' is converted to luma. Co-efficients are different between Rec. 601 and 709.

Rec. 601: Luma (Y’) = 0.299 R’ + 0.587 G’ + 0.114 B’

Rec. 709: Luma (Y’) = 0.2126 R’ + 0.7152 G’ + 0.0722 B’

The Y' component is set so that the legal range is from 16-235 (for 8-bit formats).

Cb = scale factor * (B' - Y') + offset
Cr = scale factor * (R' - Y') + offset
People mistakenly call these U and V.
I don't know all the numbers off the top of my head, but Cb and Cr have a legal range from 16 to 240.

In Vegas, Y'CbCr gets converted to "studio RGB" color space.
Studio RGB color is RGB color (not Y'CbCr), with legal range from 16-235. It does maintain superwhites, although it doesn't contain every possible Y'CbCr.

Note that Y'CbCr color space encompased a lot of very illegal colors, many of which have negative R/G/B values.
Coursedesign wrote on 9/28/2006, 9:37 AM
In FCP, you have bizarre superblack problems in that the scopes clip the superblacks (although they may or may not be clipped). No one seems to notice this.

The latest FCP 5.1.2 release supposedly fixes the scopes, I haven't had time to check.
farss wrote on 9/28/2006, 1:28 PM
I've never used FCP to know what its scopes are doing but footage out of both iMovie and FCP seems to read the same on the Vegas scopes as the camera originated material. Blacks at 0, whites peak to 110.
I still have some FCP material on disk I could check but from memory it'll leave graphics that are -10 to 110 as exactly that also, in other words both superblacks and superwhites pass through FCP the same as they do through Vegas. However it's own graphic generators (well I've only seen its text) are between 0 and 100 i.e. it generates correct levels.

Bob.