Calibrating to sRGB - Page 2

I think I may be a few steps behind you folks, but if you're trying to remap the RGB stimulus values so that...
 

My guess is that can probably be done this way...
 

(The gamma values shown in the plot in Post #25 are γ sRGB Rev. Trans. / γ Display for 0% to 100% Stimulus on a flat 2.22 display.)

To see if this works, I computed the new (ie remapped) stimulus values for a flat 2.22 display using the Stim In -> Stim Out equation above. And calculated the relative luminance of the new stimulus values to see if they match the sRGB Reverse Transform... and they were identical.
 

% Stimulus In	Gamma Correction (γ sRGB Reverse Transform / 2.2222....)	% Stimulus Out	Rel. Lum. of Stim Out on 2.22 Display (Stim Out ^ 2.2222...)
10%	.8996	12.60%	.0100
20%	.9529	21.57%	.0331
30%	.9770	30.84%	.0732
40%	.9913	40.32%	.1329
50%	1.0008	49.97%	.2140
60%	1.0078	59.76%	.3185
70%	1.0131	69.67%	.4480
80%	1.0173	79.69%	.6038
90%	1.0208	89.80%	.7874

Relative luminance values for the sRGB Reverse Transform from the table in post #5 for comparison:
 

% Stimulus	Rel. Lum. of sRGB Reverse Transform	γ sRGB Reverse Transform
10%	.0100	1.999
20%	.0331	2.118
30%	.0732	2.171
40%	.1329	2.203
50% (mid-gray)	.2140	2.224
60%	.3185	2.239
70%	.4480	2.251
80%	.6038	2.261
90%	.7874	2.268

Small reminder, the γ in equations, tables and graphs above refers to gamma, not luminance.

Edited by ADU - 2/2/13 at 1:58pm

https://ritdml.rit.edu/bitstream/han...sis04-1991.pdf

^ Looks like a fun read.

LUts in PCs may have 16bits of depth, but the actual percision is limited to the 10 most siginifigant bits. Then if you are using digital out only 8 bits go on the wire.

Well done then. Any chance of seeing the LUT for your display before the sRGB adjustments, to compare the difference?


I'm not sure which of these two approaches would be faster/easier...


It probably depends on the data you have to start with (gamma or rel. lum.). You'll probably end up in the same place though with either one.

Bump in case anyone wants to add something else before this gets archived.

Hi people.

I've always thought that sRGB was 2.2, from what I've heard. But by trying to understand this thread, you guys seem to be leaning towards 2.3? I'm calibrating to sRGB for games mainly so if I'm wrong in assuming it's 2.2, is there a right one to be choosing?

sRGB overall gamma is actually closer to 1.95...

Steve

thank you. Should I try to set gamma to that for video games?

In my experience, gamma can be different from game to game. I have my monitor set to a 2.0 gamma. Then most games have an internal adjustment that I tend to use. Some games I don't need to adjust, others I do. It also depends on what type of games you generally play, and your viewing goal. I tend to play a lot of FPS games, so I tend to set the gamma a tad on the high side for those games....I like to see the "campers" in the dark corners.

That's just flat out incorrect.

sRGB does start low with a gamma of about 1.7 at 5%, but by 10% it's at 2.0 and is inbetween 2.2-2.3 from 40% to 100%

Perhaps you should go look at the formula again.

So there is no one correct setting for gamma to be using for sRGB?

It's a formula, it has specific output for specific input.

Should people stick to 2.2 using a gamma control?

sRGB, 2.2, 2.25, 2.3 and BT.1886 are all good choices.

With software like CalMAN, it's easy to choose your calibration target.

I see thanks a lot. I'll be investing in Calman 5 soon, so it might be easier choosing.

Everyone just uses 2.2 for gamma with "sRGB" content. (i.e. computer monitors)

As I stipulated, OVERALL gamma is closer to 1.95... I never quoted the actual formula...

Yes and that's what's wrong

Overall gamma for sRGB is about 2.2, brighter down low, darker up high.

Like I said before, go check your math.

I have just re-checked the maths and it comes out at 1.956....



But, most users stick to 2.2 'cos few images on the web were ever graded for sRGB.

Yes that's why I said your math is wrong.
Something in your formula is not correct.

Go reread the standard http://www.w3.org/Graphics/Color/sRGB.html and if you care to assert that you're correct, try showing your math, because your statement is contrary to every official document on sRGB.

When you average 1-254 for PC levels you should get 2.17.
The cross over to >2.0 happens at bit 28, and the cross over to >2.2 happens at bit 102, if you need some data to check your work with.
Edited by sotti - 2/11/13 at 11:22am

Checking the maths further, but it's still coming out the same...

Will re-run the full calculations!

This might save you some trouble. And or help you check your math.

http://www.avsforum.com/t/1409045/how-power-law-gamma-calibration-can-lead-to-crushed-blacks/90#post_22208270

Download the BT1886v3.zip sheet and open it up. Checkout the Comparison of Effective “Gamma” Values" chart.
Edited by SiGGy - 2/11/13 at 11:59am

EDIT...
Edited by Light Illusion - 2/11/13 at 3:17pm

By god are you just destroying the visuals then?

An end to end gamma of 1.9 would be extremely dark.

Charles Poyton talks about using an end to end gamma of 1.1 to 1.2 in his sections on gamma in Digital Video.
http://www.poynton.com/notes/colour_and_gamma/GammaFAQ.html

The numbers we are quoting are the display gamma numbers for an end to end gamma of 1, since we are simply reversing the well known sRGB encode formaul. I'll ask again, what formulas are you using?

You seem to be out of your depth with these comments. Perhaps you should go talk to your color science guy.
Edited by sotti - 2/11/13 at 1:20pm

EDIT...
Edited by Light Illusion - 2/11/13 at 3:16pm

So clearly there is something being lost in translation.

An end to end gamma of 1, means the encoder and decoder are using inverse algorithms.

For most video content we have an end to end gamma of 1.1-1.2 because we are encoding with the rec.709 camera gamma function and decoding with a power function of 2.2-2.3

Even if what you're saying is accurate, which it's not, your comment that sRGB is really ~1.95 would still be topically wrong because we are talking specifically about display gamma, since that is the only half of end to end gamma that an end user can control.

Making contradictory statements, just to make them isn't helpful to anyone.

EDIT...
Edited by Light Illusion - 2/11/13 at 3:16pm

This comment contradicts everything in this thread up to the time you made it. Several people have pointed out the average is about 2.18, yet you claim it's 0.2 lower than that.

You dropped by, said we're wrong, won't disclose the formula you used to achieve your information, avoided addressing this by claiming you're talking about end to end gamma. Then threw your hands up and claimed your color scientist understands this, but you might not, but we're still wrong.

I gave links to the official formula from the W3, I've helped point people to the formulas used to build the BT.1886 spread sheet and implemented the formulas in CalMAN. All hard public information.

Quite true, once you start talking sense feel free to come back.

Ok, I have just been over this with our colour scientist, and no, we are not going to discuss this any further.

We have very valid reasons for the value we use, and we will not be changing it.
Very happy for others to disagree with us, as that is each and every one's prerogative.
But we stand by the value we use.

But please understand this is not something that limit our users to our values.
As we do not link profiling with calibration all users can define their own 'values' for final calibration after the profiling process has finished.

Cheers, Steve