How many colors do we need to completely cover human vision?

[mzso]

Hello!

I'm curious about this.
The familiar CIE "horseshoe" diagram implies to me that it can be done. Though I have no idea what the x/y axis represent I assume they can't go to negative, or zero.
Color systems clearly can go past the the edge of the "horseshoe" curve. As shown on the image for ProPhoto RGB.
By the looks of it the curve could be contained with a 4-5 corner (color) polygon.
https://upload.wikimedia.org/wikipedia/commons/thumb/1/1e/CIE1931xy_gamut_comparison.svg/964px-CIE1931xy_gamut_comparison.svg.png

[StainlessS]

Most humans can discern about 10 millions colors, although there are supposed to be a few that can see in the region of 100 million.

Color vision table, The table about 2/3 the way down the page.
https://en.wikipedia.org/wiki/Color_vision

Insects eg some bees and birds, eg humming birds can see in the region of 10 billion colors (some humming birds and bees can see in the Ultra Violet, and eg Rattle Snakes can see in the Infra Red [although not with their eyes, special pits approx where nostrils would be]).

[mzso]

Most humans can discern about 10 millions colors, although there are supposed to be a few that can see in the region of 100 million.

Color vision table, The table about 2/3 the way down the page.
https://en.wikipedia.org/wiki/Color_vision

Insects eg some bees and birds, eg humming birds can see in the region of 10 billion colors (some humming birds and bees can see in the Ultra Violet, and eg Rattle Snakes can see in the Infra Red [although not with their eyes, special pits approx where nostrils would be]).

Perhaps. But it's not really relevant. By mixing three colors we can cover the majority of human perception, as shown by the triangles on the picture.

[huhn]

the number of colors is irrelevant it's more about the max wavelengths.the answer to the number of colors would be 3 because the x/y can be negative check out XYZ.

[foxyshadis]

Perhaps. But it's not really relevant. By mixing three colors we can cover the majority of human perception, as shown by the triangles on the picture.

I think you're going to have to be a lot more rigorous about this question if you want a meaningful answer. Rigorously define what you mean by color and coverage, just for starters.

[poisondeathray]

I'm curious about this.
The familiar CIE "horseshoe" diagram implies to me that it can be done. Though I have no idea what the x/y axis represent I assume they can't go to negative, or zero.
Color systems clearly can go past the the edge of the "horseshoe" curve. As shown on the image for ProPhoto RGB.
By the looks of it the curve could be contained with a 4-5 corner (color) polygon.

If you're referring to the CIE chromacity diagram, ProPhoto does not have complete coverage

But yes , it can be completely covered and exceeded, by ACES for example . It has been used in actual production workflows for a few years now .

https://commons.wikimedia.org/wiki/File:CIE_xy_Chromaticity_Diagram_of_ACES_and_other_colorspaces.svg

Open Source tools like Natron can handle ACES with opencolorIO, and blender has limited support too

[mzso]

I think you're going to have to be a lot more rigorous about this question if you want a meaningful answer. Rigorously define what you mean by color and coverage, just for starters.

Well, I thought I did that particularly with the picture, where triangles are drawn for different colorspace that don't completely cover the area that represents human vision.

Anyway. What I meant was how many wavelengths/frequencies of colors (instead of RGB) to be mixed. (in a pixel for example)

[mzso]

the number of colors is irrelevant it's more about the max wavelengths.the answer to the number of colors would be 3 because the x/y can be negative check out XYZ.

If you're referring to the CIE chromacity diagram, ProPhoto does not have complete coverage

But yes , it can be completely covered and exceeded, by ACES for example . It has been used in actual production workflows for a few years now .

https://commons.wikimedia.org/wiki/File:CIE_xy_Chromaticity_Diagram_of_ACES_and_other_colorspaces.svg

Open Source tools like Natron can handle ACES with opencolorIO, and blender has limited support too

Ah, cool. Good to know.
What wavelength colors does it use? Is it still RGB? (And only more bits are required)

(Hopefully someone will update the more used image file I got from wikipedia with this colorspace)

[huhn]

there is again no answer to the bit deep question.
even 1 bit is enough if the resolution(of cause the get totally out of hands with 1 bit but not with 6 bit) is high enough and brightness has an effect on the needed bit deep too and here starts the next question what is the "maximum brightness".

XYZ is not RGB based as far as i know; https://en.wikipedia.org/wiki/CIE_1931_color_space#/media/File:CIE1931_rgxy.png

[poisondeathray]

Ah, cool. Good to know.
What wavelength colors does it use? Is it still RGB? (And only more bits are required)

(Hopefully someone will update the more used image file I got from wikipedia with this colorspace)

ACES is RGB based.

Typically 16bit half float , or 32bit full float are used . I don't know if there is a "minimum" value, but I've never seen anything ACES related below 16bit half float.

Even though the values are represented by ACES , no current display will show all of them (ie. no current display has full ACES coverage) . So right now it's just used in acquisition and post production workflows, not for end delivery

https://en.wikipedia.org/wiki/Academy_Color_Encoding_System

[mzso]

there is again no answer to the bit deep question.
even 1 bit is enough if the resolution(of cause the get totally out of hands with 1 bit but not with 6 bit) is high enough and brightness has an effect on the needed bit deep too and here starts the next question what is the "maximum brightness".

XYZ is not RGB based as far as i know; https://en.wikipedia.org/wiki/CIE_1931_color_space#/media/File:CIE1931_rgxy.png

I don't think dithering to simulate brightness levels is strictly relevant to this though.

[wonkey_monkey]

I think when asking how many bits are required, it is safe to assume that low resolution is implied (e.g. just two big pixels next to each other).

[huhn]

and again no answer to that how is the light distributed gamma, PQ, linear or some new form, how is it powered does it need processing before.

[Asmodian]

The deeper you try to understand how human vision works, especially color vision, the more "it depends" comes up. :/

The Royal Institution, The Physics and Psychology of Colour - with Andrew Hanson
https://www.youtube.com/watch?v=af78RPi6ayE

[StainlessS]

What I meant was how many wavelengths/frequencies of colors (instead of RGB) to be mixed. (in a pixel for example)

JFYI, The electro magnetic spectrum is a continuum.

Continuum
Noun: continuum (continua, continuums)
A continuous nonspatial whole or extent or succession in which no part or portion is distinct or distinguishable from adjacent parts.

[mzso]

JFYI, The electro magnetic spectrum is a continuum.

Continuum
Noun: continuum (continua, continuums)
A continuous nonspatial whole or extent or succession in which no part or portion is distinct or distinguishable from adjacent parts.

True. But human vision is not equipped to discern arbitrary frequencies. It can be (apparently completely) fooled by three colors of a particular frequency.

[huhn]

the AVG human as 3 cones and the brain and some cells uses a lot of math to do the "rest".

[wonkey_monkey]

True. But human vision is not equipped to discern arbitrary frequencies. It can be (apparently completely) fooled by three colors of a particular frequency.

Not completely, and certainly not by using the peak cone response frequencies. And practicalities mean that the typical colours used in displays are a compromise. You can't replicate the deep violets you see in a sunset on a typical RGB display, for example.

[mzso]

Not completely, .

Well, that's not what the triangle for the ACES colorspace suggests. (Image linked above)

And practicalities mean that the typical colours used in displays are a compromise. You can't replicate the deep violets you see in a sunset on a typical RGB display, for example.

This is of course true, for the time being.

[Asmodian]

Lasers are great for exploring the limits of color vision, you can get a lot of pure colors that you cannot represent with any normal display (or normal objects). You don't usually see yellows or blues that color. A 473 nm laser is a cyan that is way outside what a normal display can do. It looks weird. :p