<!>Maps (2014-10-02 16:23:43)
Maps
Anthologica Universe Atlas / Forums / Miscellaneria / Maps / <!>Maps (2014-10-02 16:23:43)

? Rhetorica Your Writing System Sucks
posts: 1292
, Kelatetía of Space
message
The three attributes of colour that people usually consider when designing something for visual appeal are hue, saturation, and value/luminosity; value is essentially brightness (lighter colours vs. black), saturation is essentially colourfulness (strong colours vs. grey) and hue is the average wavelength of the colour as perceived by the eye (i.e. a rainbow.) Luminosity is this weird in-between thing between value and saturation where the range is extended upwards towards white, and the actual bright colour itself is in the middle. Windows uses an HSL colour picker by default, whereas Photoshop uses HSV.

hsv_hsl.png

Due to boring yet fun technical details, however, the human eye doesn't actually work this way; it detects hues as a pair of opposing poles that it contrasts, and grey is simply the occurrence of neutral values in both channels:

about-img1.jpg

(Unimaginatively these poles are called 'a' and 'b', which don't stand for anything. This is called Lab colourspace.)

Computers, of course, use the RGB colourspace for display:

RGB_Cube_Show_lowgamma_cutout_b.png

This has interesting consequences because (standard) human eyes are more sensitive to green than red and more sensitive to red than blue. When converting colour images into greyscale, the final value is generally generated with the formula:

v = 0.2989 * r + 0.5870 * g + 0.1140 * b

...which should give you some sense as to the relative brightnesses of the hues in question. There's been a lot of work in recent years on quality of colour selection in data visualization, chiefly as a result of awareness about how this stuff works. For example, if you want to make a heatmap (where different colour schemes correspond to different tiers of activity) you need to make sure your gradient evenly distributes luminosity according to human perception, or the eye will be misled by subtle imbalances in the scale. The most prominent project designed to improve this sort of thing are the Brewer palettes, which are built in the LCH space, a variant of HSL that was built to correct for peculiarities in human perception.

The Brewer palettes also include rainbows that you can use to ensure maximum visibility despite colour vision impairments, and they're honestly pretty attractive, too.

Critically, no two colourspaces are created equal: they each have areas that they emphasize better (e.g. HSV and RGB have more detail for greys than Lab), and none of them are expressive enough to describe everything that a human eye can see, excepting perhaps those with severely deficient monochromatic vision. This, however, is largely because the elements in computer displays do not accurately match the response frequencies in the human eye's receptors:

Eye:

Human_spectral_sensitivity_small.jpg

LED:

Fig-2-white-LED-vs-RGB-LED-spectrum.jpg

CRT:

CRT_phosphors.png

...The key issue, however, is that the human eye determines hue through subtractive combination, a process meant to account for the possibility of mutational drift in the response curve of the actual cone receptors as a result of evolution. This is a computationally and instrumentally difficult process to accommodate, and no one really wants to do the necessary math to make a display that accurately reflects this aspect of human vision, so we use direct linear combinations of the three RGB elements as a "good enough" compromise. Still, there have been promising results in emulating nature by adding elements to display more of the spectrum rather than attempting to make something that is specifically tied into expectations about perception. (Although, from a conworlder's perspective, we're still biased, since the new channel targets the area of the spectrum that is densest for us boring ol' mammals.)