are all bright pixels equally bright?

Some colors like #FFFF00, #00FFFF, #00FFFF, #FF00FF, #00FF00 seem really bright while others like #FF9900, #2D4FFF, #6699FF,#4603FF, #00FF5C don't.

I ask because I thought that any color with at least one maxed component forces the pixel to be at maximum brightness, so I would imagine every color with an #ff in it should be equally bright on screen.

Is this purely my perception, or are these colors quantifiably less bright than others on my screen?

• The pixel is as bright as it can be, but not all colors are equally bright. Pure red, for example, is brighter than pure blue. Does that answer your question? Commented Jul 18, 2018 at 23:03
• i wanted more detail on how to tell whether a pixel cluster is brighter or darker than another, how do you know pure red is brighter than pure blue? Commented Jul 18, 2018 at 23:06
• Because Wikipedia tells me (not that I needed that). The PAL conversion from RGB to gray is 0.39*R + 0.40*G + 0.11*B, so a maximized Blue value cannot physically be as bright as a maximum Red or Green. Commented Jul 18, 2018 at 23:09
• oh so pal is a better indicator of brightness than brightest component of rgb? Commented Jul 18, 2018 at 23:10
• @Dmitry its not pal, its just one possible formula for the relative sensitivity of human eye. Even if the colors would be equally bright does not mean your eye records it that way. Commented Jul 19, 2018 at 7:44

Your problem pops up regularly. That's because RGB numbers do not properly present perceived luminosity. More confusion is caused by hue-saturation-brightness(or luminosity) presentation of RGB numbers. That's only a math transform. Perceived luminosity doesn't follow only the brightness component, hue and saturation affect very much, too.

If somebody has designed how our sight work, he at least hasn't had any respect on our RGB computer color systems.

The problem is well known and several attempts has been done to present colors in computers so, that the numbers present the peceived values more accurately.

Different software support differently those attempts. In Photoshop we have Lab color mode. GIMP has HCL, which seems to be a polar coordinate version of Lab. Krita puts it further. There's even XYZ, the original vision modelling based predecessor of RGB.

If you want make subjectively as bright colors starting from a set of RGB colors, convert them to Lab in Photoshop and turn off the lightness channel in the channels panel.

It's done here to your example pattern. Test yourself and try to guess where the white spot has vanished.

(beware, this screenshot isn't especially accurate because there's numerous conversions between your original and the attachment)

Do not expect these colors are usable by simply returning back to RGB mode and picking the RGB numbers. Probably many of them are clipped due the limited color range of RGB - they are false onscreen and give another way false results when returning back to RGB mode.

If you put on Gamut warning and proof color display and select for sRGB proof colors, many spots are greyed, which means out of gamut.

Proof setup:

(relative rendering in the displayable range)

You can compress the color range to displayable following way:

1. Set all to same lightness with Curves:

1. Reduce the colorfulness with saturation slider in Hue&Saturation dialog

Now the spots stay quite same after returning to RGB mode:

How much they differ, it depends on how well your system is calibrated.

You should be careful before you extend mathematical intuition on colors. There are just so many false assumptions you can make.

First, you have be a bit careful with your nomenclature. While it is normal to call a RGB triplet a color, it is not. It is a device specific instruction that produces a different color on each device*. Additionally intensity may mean how many photons are emitted per unit of time or how intense humans percieve them.

Second, human vision nor standard RGB colorspaces are not linear. This means that #111111 is not half as bright as #22222 even though the numbers are. The separate channels are independent of each other, or as separate as possible for human vision. This means that #FF0000 is less bright than #FF2200, both physically and perceptully. But the channels dont have the same relative intensity so #FF0000 is not same intensity as #00FF00**. (yes this means that the blue of the sky is way brighter than green vegetation even when the appeaar as bright)

So how do you do this. Well it depends on how accurate you want to be. If possible you would use something like CIE Lab, and rely on your Color management system. If your not that particular then you can use any of tge ballpark polar solutions like HSB or HSL see vikipedia for how to compute these.

But in reality the whole story is quite complicated, we didnt even get started on human white balancing. Which is why you get the blue or gold dress discussion.

*That is unless you have a recently profiled/calibrated device and your imaging system is using a color manager.

** and this is where the PAL formula comes in as a common approximation.

• can you explain in more detail why #111111 is not half of #222222? I know that #000000 is not actually black hole black, and #ffffff on a screen may as well be black relative to a surrounding that is significantly brighter, but I am not sure why #111111 is not half of #222222 in terms of red blue and green intensity. Commented Jul 19, 2018 at 10:26
• @Dmitry thats simply what not linear or nonlinear means. If the color wasnt stored as nonlinear then it would mean that RGB would store a lot of colors that humans could see no difference between and in the other end miss a lot o nuance. Simply because human eyes are not equally sensive to bright and dark colors. Commented Jul 19, 2018 at 10:44

Interesting question but, with some faulty assumptions.

The human eye is not equally sensitive to all colors.

The responsiveness of out eye starts from not perceiving a wavelength at all, below ultraviolet, starting to perceive blues, greens, reds, and the sensitivity starts do decay next to infrared.

Here is a fake graph.

But our eyes because the fisiology behind it can be fooled by using just some combination of narrow bands.

If the eyes work as you expected would look like this

Here I made a color chart based on HSB, this is Hue, Brightness, and Saturation.

If I extract the brightness channel, I have what I expected to be when I made this chart in the first palace. All colors in the now disappeared white circle have in fact at least one component #FF.

But as the human eye is neither mathematical or theoretical, we do not perceive as this.

If I want to emulate the perceived brightness, I need to simulate the responsiveness of the eye, and I get something like this.

Which have sense, in the center, we have photons firing from all 3 led colors, therefore it is brighter.

We are more sensitive to Green, followed by green, and last from blue. So Green (bright) and red (bright) together gives yellow perceived as the brighter angle of the wheel.

That is why there are a lot of different 3D models to represent color, RGB cubes, Color wheels, Color cones, Double cones, spheres, amorphic spheres as the Munsell bubble, etc.

And each of them has a different purpose.

There are different meanings here. All colors on the 4th image have the exact same Brightness value of the pixels. This brightness is given by the max value of the pixel. There are no subtle differences there.

But that is only because someone on that specific COLOR MODEL made it that way.

But that is not a Perceptual color model. A perceptual color model would be the Munsell "Buble", but as each human perceives things differently it cannot be a mathematical universal color model. Some people have different types of daltonism, for example. That is to illustrate that each of us could potentially see "brightness" differently. This can also apply to the viewing conditions.

Or eyes can also be "tired" of seeing one color, so this brightness varies inclusive in our own eyes.

I'll post an example of this.

Cover one eye with one hand now. Don't ask, just do it and keep reading!

Now stare at the red rectangle for some 20seconds. After that, scroll down fast and alternate viewing the ball's image with your left and right eye, in rapid succession and see the differences.

I now broke one of your eyes and de-calibrated it.

You can not rely on a perceptual model...

Original ball image

• so, there are two problems, both that objectively pixels with an ff component do realistically very in brightness, subtle brightness differences from this can be amplified/diminished by our color sensitivity, and that after that, we have to deal with the color sensitivity for the resulting color as well. Commented Jul 19, 2018 at 18:17
• I complimented my answer based on your comment. Commented Jul 19, 2018 at 18:33
• no hotlinking :/ Commented Jul 19, 2018 at 18:36
• @Dmitry Human visual system is mindboglingly complicated. You could spend a full masters degree on the subject and still not ready. Commented Jul 19, 2018 at 18:59
• A hsl wheel tries to accoint for the differences in lumniosity of each channel. As does Lab Commented Jul 19, 2018 at 19:01