Black & White vs Desaturation

robwood wrote on 10/12/2011, 10:48 AM
Ok... this has bugged me for YEARS but today I felt like displaying my confusion

Why doesn't the Black & White filter work the same as using Saturation=0% in a CC filter?

I always see luma differences up to 20% whenever I compare them... yet in the help it says "This filter removes all color information from an image without altering the luminosity." So I never use the B/W filter...

many ways to set up a compare, my fave is difference mode;
identical image on two layers, 1 filter desat, 1 filter b/w
set top track to dff, any return other than BLACK means difference, scopes will tell how much difference

I use CC all the time anyway, so it seems like a non-issue ("don't use B/W, use desat"), but sometimes I wonder about it.

Comments

malowz wrote on 10/12/2011, 11:31 AM
as you already said, saturation reduce the saturation channel in HSL. (hue, saturation, luminosity)

but saturation increases luminosity, cause colors have their luminosity level.

when using black/white function, it knows how much light a color add, and preserves this luminosity. reducing saturation does not do that.

there is no "wright" way, B/W can be the most common or "standard", but as most people like the "artistic" of black and white, any advanced/non-standard way can be better sometimes ;)

also, Black and White can have a enhanced contrast by using "channel blend" filter, setting more luminosity from each primaries, enhancing "visual appealing"

http://www.cambridgeincolour.com/tutorials/color-black-white.htm
robwood wrote on 10/12/2011, 11:56 AM
that sounds great, but my scopes say the luma values are being altered when i use black and white, but not altered when desaturating... shouldn't the reverse be the case if what you're saying is correct?
GlennChan wrote on 10/12/2011, 1:13 PM
There are different formulas to convert RGB into other colorspaces, the other colorspaces being:

HSL
Y'CbCr with Rec. 601 coefficients <-- This is what the 3-way color corrector does
Y'CbCr with Rec. 709 coefficients
and others

Vegas' scopes will show LUMA (Y') with Rec. 601 coefficients. Luma is NOT the same thing as the scientific definition of luminance.

In the HSL color model, the L stands for luminosity and has nothing to do with the scientific definition of luminosity, nor does it have anything to do with luma or luminance.

You can look up the formulas online. "Saturation" is defined very differently in these color models. / The formulas used are very different.
robwood wrote on 10/12/2011, 3:44 PM
"Y'CbCr with Rec. 601 coefficients <-- This is what the 3-way color corrector does" - Glenn

The shift in the midpoint is quite noticeable when using B&W filter (waveform midpoint drops 5 to 10 IRE).

Are there specific circumstances when the B&W filter is more appropriate than using Desat in the 3-way?
Or is there no significant difference other than the midpoint drop?

(i'm guessing the second as I've never read anyone citing specific uses for B&W filter in Vegas over the CC's)

GlennChan wrote on 10/13/2011, 1:19 AM
Well the picture will be slightly different.

It basically comes down to what you want the image to look like. See:
1- the cambridge in color link earlier in this thread
2- http://www.glennchan.info/fcpugto/shot10-%20Channel%20Blend/shot10.htm
Grazie wrote on 10/13/2011, 1:28 AM
Glenn "[I]It basically comes down to what you want the image to look like.[/I]"

And with Glenn's monumental understanding, of what goes on under the hood/bonnet, gives us the ability/tools/understanding to choose where to go next. Personally, when he talks about these elegant matters, I'm in complete awe at his knowledge. Having said that, I'm going to nail his quote above, to my editing screen!

Grazie

Rory Cooper wrote on 10/13/2011, 1:49 AM
That is a good question, regarding the technical scientific aspect I don’t have a clue.

for me personally the differences would be applied in obtaining cleaner harder masks or deeper gradient masks, for example in this clip the differences are noticeable on the greens as in Chroma key green. The other tones are hardly affected. So it has artistic and masking value but that’s it.




How I understand it that the luma values for Rec. 601 & Rec. 709 are so close it’s not worth taking into account, the real differences are in the colorspaces, regarding broadcasting these differences are not respected at all, we have been living with these differences for years. The guys who are respecting these differences are the guys producing HD content for DVD. Hopefully Glen will help me out and give me a better understanding.
robwood wrote on 10/13/2011, 10:57 AM
www.glennchan.info/fcpugto/shot10-%20Channel%20Blend/shot10.htm

thx Glenn, that page looks like it's discussing what i was seeing.
i'll do a read-over when i take a break...


LATER

...ok, read your article; hadn't thought of using Channel Blend "greyscale" setting, so tried it and compared to Black & White filter; and yep, the luma values changed the same way Black & White filter did. So guess that's how Black & White arrived at the result it did...still don't get Channel Blend as a filter, but that'll be another adventure.


"I'm in complete awe at his knowledge." - Grazie

+1 Yep. Thx again Glenn.
GlennChan wrote on 10/13/2011, 8:22 PM
Having said that, I'm going to nail his quote above, to my editing screen!

Hehehe, sometimes you don't really need to know what's going on. Making the image black and white is clearly wrong from a technical perspective. ;)

You could also try adding a slight tone to the final black and white image. In Sin City for example, the highlights are slightly yellow while the shadows are slightly blue. If you are going for a sepia tone, it helps to drag the 3-way Color Corrector to slightly different shades.

How I understand it that the luma values for Rec. 601 & Rec. 709 are so close it’s not worth taking into account
In the context of chroma subsampling, the differences between the Rec. 601 and 709 luma coefficients isn't really noticeable in real world situations.

see this article
http://www.glennchan.info/articles/technical/chromata/chromata.html
scroll down to "Luma/Luminance Co-efficients- Rec. 601 verus Rec. 709 " for a demonstration.
*There are some inaccuracies in that article.

If you are doing a Y'CbCr based de-saturation... then there will be a subtle difference depending on which luma co-efficients are being used.
Y' = 0.299 R' + 0.587 G' + 0.114 B' (601 coefficients)
Y' = 0.2126 R' + 0.7152 G' + 0.0722 B' (709 coefficients)

As a point of comparison, if you use a "red only" black and white conversion, then the equivalent formula would be
Y' = 1.000 R'

So the luma coefficients are kind of a play on how much R'/G'/B' contribute to the black and white conversion.

In practice, you probably don't care.
Grazie wrote on 10/13/2011, 9:29 PM
Glenn, I liked what you said about Sin City. Doing this to a visual is kinda like applying some ever so slight reverb to audio JUST to add some texture/sweetener to the sound coming out of the speakers.

Grazie