RGB luminosity

Graphics and GPU Programming Programming

Started by Jason Doucette February 11, 2005 09:34 PM

17 comments, last by Dmytry 19 years, 2 months ago

1,806

February 12, 2005 06:50 AM

Quote:Original post by Dmytry
and other posts, after that I thought you were talking about visible lightness, and that values looked rather like responses of eye (most sensitive to green).

The linear equation Y = a*R + b*G + c*B is radiometric per-se. Rec 709 et al try to adjust the factors a,b and c in order to accomodate the visual response curve for the three monochromatic primaries, basically making it photometric. But this is just a rough approximation, as I have mentioned several times. If you want good precision, then you need to use the combination of ICC profile and CIE weighting. Still, the decimals are not "speculation", they're the mathematical closed form solution of the psycho visual model, when applied to three monochromatic primaries. Keep in mind that Rec 709 wasn't designed to compute the perceived intensity (that's what the CIE curves are for), but it was designed to specify electrical signal levels for HDTV.

Quote:Original post by Dmytry
If that is radiometric equation, then it doesn't take human eye into account, and should not take human eye into account, and therefore does not depend to human, or to color blindness/weakness.

I said that there were special CIE curves for colour weaknesses - that's a difference, since they are in fact non-linear.

Quote:Original post by Dmytry
And yes, then it could be precise.(but, certanly not to fifth digit (heck, if you'll tell somebody who really works with colors that displays have precision to fifth digit, he will just laugh. It at best have +/- 1% )

Keep in mind, that these equations are not only used for your typical CRT display. They are just as well used for eg. laser imaging devices, which can in fact be much more precise than 1%. It is important to keep the precision, in order to make the equations usable on any device and image generation technology.

Dmytry

1,151

February 12, 2005 07:30 AM

I know that this values is solution of some equation. But all coefficients in equations is known with certain precision. Really, if my height is 180 cm +/- 5mm , and i convert it to inches, if i keep fifth digit precision, that digits will be just speculation anyway, right? Created to mislead people that it is something uber precise. Same there. We have some equation where almost all coefficients are +/- some precents. And then, we magically get five-digit precise result from it, heck.

Do you know *what* really is fifth-digit precision? It's 0.000008 cm^2 effective area of dirt on 1 cm big lens, that's it. It is extremely hard (if possible) to, e.g. measure electrical current with that precision. For example, i know that projection systems inherently can't have that precision because opacity of air depends to themperature.[grin][lol]

edit: and laser imaging devices have completely different spectrum.

Yann L

1,806

February 12, 2005 07:49 AM

Quote:Original post by Dmytry
I know that this values is solution of some equation. But all coefficients in equations is known with certain precision. Really, if my height is 180 cm +/- 5mm , and i convert it to inches, if i keep fifth digit precision, that digits will be just speculation anyway, right? Created to mislead people that it is something uber precise. Same there. We have some equation where almost all coefficients are +/- some precents. And then, we magically get five-digit precise result from it, heck.

Not in this case. All involved quantities (except for the psycho visual curves, which are fuzzy by nature) are defined well beyond the fifth decimal.

Quote:Original post by Dmytry
Do you know *what* really is fifth-digit precision? It's 0.000008 cm^2 effective area of dirt on 1 cm big lens, that's it. It is extremely hard (if possible) to, e.g. measure electrical current with that precision.

Electrical current can be measured down to the single electron. Although this is obviously irrelevant for this specific application, one have to take into account, that many radiometric equations have applications far beyond computer imaging. There are many circumstances where such precision is required (for example in high energy and particle physics simulation software).

Quote:Original post by Dmytry
For example, i know that projection systems inherently can't have that precision because opacity of air depends to themperature.[grin][lol]

You do realize, that certain extensions to ICC profiles (for example for very high fidelity image beamers) can take environmental parameters such as air temperature, pressure, dust density, etc, into account ? Those are important to minimize chromatic abberations, when projecting on very large surfaces, especially outside.

Quote:Original post by Dmytry
edit: and laser imaging devices have completely different spectrum.

Irrelevant. That's what ICC profiles are for - to standarize the chromaticities of devices with totally incompatible spectra. Just look at printers, they don't even have the same colourspace.

BTW, this is going heavily offtopic, and we should stop this discussion. If you're interested in continuing it, I'd suggest opening a new thread.

Dmytry

1,151

February 12, 2005 07:59 AM

Quote:
Not in this case. All involved quantities (except for the psycho visual curves, which are fuzzy by nature) are defined well beyond the fifth decimal.

And then, we have well-known "average precision" sindrome[grin]. Do you know that if one thing is inprecise (psycho visual curves), result is as inprecise?

(as about "well beyond the fifth decimal" i doubt even single value is measured with that precision, but anyway.)

Quote:
Electrical current can be measured down to the single electron.

Electrical charge (not current) can be measured down to single electron. You can measure average *very small* current during some relatively big amount of time (seconds) with that precision, by counting every electron. But nobody really does it, and it is practically irrelevant.
And electron counters, too, could make mistakes (miss electron, or count non-existing electron) much more often than 1 electoron per 100000 .
And almost every automated measurement boils down to measurement of current. It is definately possible to count photons, or electrons, but it can't be done fast or reliably. It is VERY hard to measure ampers of current with five-digit precision, or even microampers.

Quote:
There are many circumstances where such precision is required (for example in high energy and particle physics simulation software).

Exactly. In that areas such precision is required.

I have no idea what argumentation could be used against man deeply believing that vision systems could have 5-digit precision of colors. Eye doesn't have that precision, damn. No human sense except counting have such precision.

Quote:
You do realize, that certain extensions to ICC profiles (for example for very high fidelity image beamers) can take environmental parameters such as air temperature, pressure, dust density, etc, into account ? Those are important to minimize chromatic abberations, when projecting on very large surfaces, especially outside.

Of course, with large surfaces, you have to take that into account because it's affects result by _several_precents_.

[Edited by - Dmytry on February 12, 2005 8:59:06 AM]

Yann L

1,806

February 12, 2005 09:50 AM

Well, this is definitely going too far offtopic.

Anyway, whether you like it or not, those values are internationally accepted standards. If you use different values, or truncate them, then your results will not be comparable or reproduceable by any other standarized system on the planet. There is a reason for the standarization of such conversion weights, and I would strongly recommend to use them.

Oluseyi

2,123

February 12, 2005 09:56 AM

Thanks to both of you. I learned a lot, probably more than I intended to!

Raghar

February 12, 2005 06:56 PM

Dymitry

Quote:
everything except "0.2 0.7 0.07" is pure speculation

You'd have problems if you'd use 0.2 0.7 0.07.
0.2 + 0.7 + a MUST equal 1. In your example you'd have darker image than original.

I would have another problem. It's difference between CRT and LCD monitors. Do anyone have an experience with 24bit colors vs color depth used on LCD monitors?

Yann L

1,806

February 13, 2005 12:26 AM

Quote:Original post by Raghar
I would have another problem. It's difference between CRT and LCD monitors. Do anyone have an experience with 24bit colors vs color depth used on LCD monitors?

Your displays' ICC profile should take care of all that.

Dmytry

1,151

February 13, 2005 03:22 AM

Quote:Original post by Raghar
Dymitry
Quote:
everything except "0.2 0.7 0.07" is pure speculation

You'd have problems if you'd use 0.2 0.7 0.07.
0.2 + 0.7 + a MUST equal 1. In your example you'd have darker image than original.

ops. actually i assumed it is "renormalized" anyway (divided by sum of 'em)... then, 0.22,0.71,0.07

RGB luminosity

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

RGB luminosity

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Reticulating splines