Gamma correction

[Nicolas Lossec]

Hi, i've got few questions about gamma correction. I have the feeling that i could be totally wrong

1- In Resolve (for example), when we set the working space at 2.4 gamma, this means that the signal is actually a power of 1/2.4, right ?
We're working on the inverse function, because the monitor is doing "the real"  gamma correction at ^2.4
So in scopes, what we are seeing is also the signal with a 1/2.4=0.42 gamma !?
Can you confirm that ?

2- Do we need to make a gamma compensation when exporting a sRGB file from a BT1886 master ?
If i understand it correctly, gammas are here (at least at the end of the pipeline) to compensate the differents viewing environments. sRGB is made for a bright environment, so the image should be brighter than BT1886, isn't it ? Then :

- If i just play the BT1886 master (roughly gamma 1/2.4) on a computer, the monitor applies a gamma correction of 2.2 : the image is brighter but appears ok because there is a lot of surrounding light.

- If i export my master with a color transformation node from 2.4 to 2.2, the image should be the same on the computer than on my reference monitor, but appears too dark in the bright environment.

What should i do ?
 

[Igor Riđanović]

Yes, the display set for 2.4 gamma will require the incoming signal to have inverse gamma of 1/2.4.

Here's more from Poynton: http://sites.biology.duke.edu/johnsenlab/pdfs/tech/gammaFAQ.pdf

[Mazze]

Just curious, what monitor you are driving here?

I would think, that you could get away by just setting Resolve to output a linear image then.

Set the input gamma to 2.4 and the output to linear (which should apply the inverse gamma function and hence create a linear (or gamma 1) image), which your monitor seems to expect.

Usually you create different deliverables for broadcast and internet use - like you pointed out, one with 2.2 gamma and one with 2.4 gamma.

 

[Nicolas Lossec]

2 hours ago, Mazze said:

Usually you create different deliverables for broadcast and internet use - like you pointed out, one with 2.2 gamma and one with 2.4 gamma.

Yes, i'm sure it's better to do a trim pass manually to make those deliverables.
But is there a way to do it automatically ?

 

[Andy Minuth]

I don't do different passes for sRGB. Client gets just one Master in Rec.1886. One reason You already mentioned: because the surround is part of a color-space (bright vs dim in this case).

The other point is: How do You know that the client will not upload his "Video ProRes" to YouTube? And when it is on YouTube, what if someone watches it on their Television, etc. You can go on like that...

From my point of view, the player should do the correct gamma correction for the display. So I try to just give proper Rec.1886 masters with correct metadata.

[Orash Rahnema]

1 hour ago, Andy Minuth said:

I don't do different passes for sRGB. Client gets just one Master in Rec.1886. One reason You already mentioned: because the surround is part of a color-space (bright vs dim in this case).

The other point is: How do You know that the client will not upload his "Video ProRes" to YouTube? And when it is on YouTube, what if someone watches it on their Television, etc. You can go on like that...

From my point of view, the player should do the correct gamma correction for the display. So I try to just give proper Rec.1886 masters with correct metadata.

That's interesting, I always try to do the 2 deliverables but i actually never thought  about people watching youtube and vimeo on tv.

True that the surround is part of the colour space, but wasn't that the reason why computer screen manufacturer choose a 2.2. gamma?
To compensate for a brighter surround? 

How do you attach a metadata to a prores or an h264 in order to let the player do the correct conversion?

Thanks!

Orash

 

[Orash Rahnema]

11 hours ago, Nicolas Lossec said:

Yes, i'm sure it's better to do a trim pass manually to make those deliverables.
But is there a way to do it automatically ?

 

In davinci I use a color space transform ofx on the timeline level.

Setting color space and gamma to be converted from where you are to where you need to go.

[Andy Minuth]

On 09/08/2017 at 0:27 PM, Orash Rahnema said:

True that the surround is part of the colour space, but wasn't that the reason why computer screen manufacturer choose a 2.2. gamma?
To compensate for a brighter surround? 

That was my point, that a gamma correction might not be necessary. If it becomes necessary it can still be done in the playback path or the display. For example by calibrating a display to Rec.1886 or sRGB depending on the surround. 

[Nicolas Lossec]

19 minutes ago, Andy Minuth said:

That was my point, that a gamma correction might not be necessary.

I agree, even without any software ajustment.
I've done some tests and here are my results (what works for me) :

-The sRGB function in Color Transform is unusable, it crushes the blacks too much.

-From BT1886 to sRGB : a 2.4 to 2.2 correction is not necessary. A computer screen shows the original 1/2.4 gamma a bit brighter, which is nice in a bright environment.

-From BT1866 to DCP : a 2.4 to 2.6 correction is not wanted, as i want the DCP to be darker in theater. For a fast delivery, i will only do the Rec709 to XYZ transform, but i really think that a trim pass is needed or at least a proper "Output Device Transform" in ACES.

[Steve Shaw]

BT1886 is just a display calibration gamma, and should not be used for any 'conversions'. You should only use Power Law gamma for conversions.

Also, Rec709 is 2.4 gamma, and sRGB 2.2.

A display with a zero black will calibrate using BT1886 to a 2.4 power law. With a raised black you will get lifted shadows. This is one of the issues with BT1886, as the lifted shadows will cause you, the colourist. to over compensate by grading the shadows darker. The end result will be potentially crushed shadows/blacks when viewed on a display that is not calibrated to Bt1886, but has a lifted black point, as many home TV have...

Steve