This is why I much prefer the Lab color space over HSV or HSL— it's much better suited to interpolation and mixing.
The critical properties are:
* It's a vector space, so mixing colors means mathematically mixing the components. There is no polar angle that can "go around the wrong way", or suddenly cause the interpolation to change when one color coordinate crosses zero degrees.
* It's perceptually uniform, so that taking mathematically even steps in color space means the result will be perceptually even as well (compare to color spaces which don't account for gamma, and small change on a dark color make a huge difference, but the same step in a light area is invisible).
* Because of the perceptual uniformity, there are no luminance dips or peaks in interpolated colors. In RGB, for example, interpolating two saturated colors which are far away from each other on the color wheel will result in dark grey or black in the middle of the gradient— not so with Lab. Also, consider an HSV rainbow: If you desaturate it, it isn't a uniform grey– some parts of the wheel are randomly brighter than others, because the human eye is more sensitive to yellow and green light, and very insensitive to blue. A rainbow with constant L coordinate in Lab space will have constant luminance.
* Because chromaticity is a 2D vector space in Lab color (rather than a single coordinate), interpolating between two distant hues won't put crazy unrelated hues in between them, but HSL and HSV will. Saturated, complementary colors will appropriately interpolate through a neutral grey in Lab, instead of through a "random" saturated hue.
Here's [1] an image of some constant-luminance (i.e., constant "L" coordinate) slices of Lab space, to give a sense of what it looks like.
Here's [2] a little color picker (found with a Google search) which shows you the results of interpolating in different spaces— you can see that RGB is pretty bad, but in HSV or HSL it's not hard to make something look extremely broken. Those two spaces are definitely the wrong choice for interpolation.
As an artist and designer I somehow can't get over the complementary hues in the Lab color space, NCS complementaries fit much better visually.
The main advantage of mathematically derived and absolutete color models like Lab is that they can't be protected by copyright. There is an intiative that tries to be an alternative to Pantone in that market: https://www.freiefarbe.de/en/
But is a perceptual colorspace "natural"? A physics-based vectorspace is the most natural thing. Adding vectors in a perception based vectorspace seems unnatural to me.
Our human colour perception is very much not based on the physics or mathematics of light [1] [2]. It evolved to keep us alive and breeding, not to be accurate or precise about what light and colour actually are, so what is "natural" really means how we in particular see. For anything intended to be viewed by people, perception is all that matters.
Our computers, phones, TVs, etc. are designed specifically for human eyes, and other animals looking at them won't necessarily see the things portrayed with the same colours they see in reality.
Lab is not energy linear, and if you're doing rendering/illumination, you won't want to use Lab (unless you're deliberately trying to stylize). It is designed to be a good model of human color perception, not the physics of light.
Lab color space is good for mixing colors in a way that behaves intuitively. E.g. "what color is 'halfway between' these two other colors?", mixing "paint" digitally, producing pleasing gradients, picking colors or creating palettes.
The critical properties are:
* It's a vector space, so mixing colors means mathematically mixing the components. There is no polar angle that can "go around the wrong way", or suddenly cause the interpolation to change when one color coordinate crosses zero degrees.
* It's perceptually uniform, so that taking mathematically even steps in color space means the result will be perceptually even as well (compare to color spaces which don't account for gamma, and small change on a dark color make a huge difference, but the same step in a light area is invisible).
* Because of the perceptual uniformity, there are no luminance dips or peaks in interpolated colors. In RGB, for example, interpolating two saturated colors which are far away from each other on the color wheel will result in dark grey or black in the middle of the gradient— not so with Lab. Also, consider an HSV rainbow: If you desaturate it, it isn't a uniform grey– some parts of the wheel are randomly brighter than others, because the human eye is more sensitive to yellow and green light, and very insensitive to blue. A rainbow with constant L coordinate in Lab space will have constant luminance.
* Because chromaticity is a 2D vector space in Lab color (rather than a single coordinate), interpolating between two distant hues won't put crazy unrelated hues in between them, but HSL and HSV will. Saturated, complementary colors will appropriately interpolate through a neutral grey in Lab, instead of through a "random" saturated hue.
Here's [1] an image of some constant-luminance (i.e., constant "L" coordinate) slices of Lab space, to give a sense of what it looks like.
Here's [2] a little color picker (found with a Google search) which shows you the results of interpolating in different spaces— you can see that RGB is pretty bad, but in HSV or HSL it's not hard to make something look extremely broken. Those two spaces are definitely the wrong choice for interpolation.
[1] https://en.wikipedia.org/wiki/CIELAB_color_space#/media/File... [2] http://davidjohnstone.net/pages/lch-lab-colour-gradient-pick...