Hacker News Reader

Comment by __david__ | Hacker News Reader

He says "value", but he seems to mean "lightness". See http://en.wikipedia.org/wiki/HSL_and_HSV for a discussion (and neat graphs) of the differences between the two. I really like using HSL to choose colors since it is very intuitive (much more than RGB) and it's really nice to programmatically choose similar colors by just tweaking the saturation and lightness parameters.

View on HN

First, value and lightness are synonyms in this context. The name value comes from the Munsell system, http://en.wikipedia.org/wiki/Munsell near the beginning of the 20th century, and the name lightness is what color scientists finally settled on a few decades later. Unfortunately the terms’ usage in the HSL and HSV models doesn’t have terribly much to do with the human-vision-relevant definition.

It’s too bad that the author of the linked-to post claims that chroma and saturation mean the same thing though. They don’t.

Anyhow, thanks, I’m glad you liked the pictures (I drew many of them). If there’s anything about that HSL/HSV article that could use improvement, feel free to comment here or send me an email (or you could of course leave a message on the wiki talk page).

For what it’s worth, I don’t think HSL is very intuitive. It’s basically the same thing as RGB, just tilted on one end and squished a bit into a cylinder shape (as shown here e.g., http://en.wikipedia.org/wiki/File:Hsl-and-hsv.svg). For some discussion of why, see http://en.wikipedia.org/wiki/HSL_and_HSV#Disadvantages but the short explanation is: HSL is a model designed for the hardware limitations of 1970s workstations, and much more perceptually relevant models are computationally cheap enough today.

The idea would be that human vision is more responsive to ligh intensity than color intensity, and so, to optimize a display medium, we can use more bits for lightness and less for other color information. And to do that, we need a color space that treats lightness as a separate coordinate.

Secondly, designers need a color space that conforms more naturally to human vision.

I think we should be doing most of this math in floating point, so it's not about optimizing numbers of bits (modern buses are getting fast enough to send 12 or 16 bits per channel per pixel to the display, so we don’t really need to optimize bandwidth there either necessarily), but mostly about helping everyone (designers, photographers, regular people) use models which are related to the way we see, rather than computer reproduction technology.

We could attempt to use an "unbound" version of a color sistem (i.e. RGB with -inf..+inf ranges rather than 0-255 for each channel). Even if the display medium can only handle a limited band of colors, more color information may be useful (like in HDR).

I find that HSL is intuitive when I'm trying to get some similar colors together for a UI element where I need kind of a gray version and a highlight version in addition to the base color. The grey is just saturation dropped and the highlight is some extra lightness. If the hue stays the same then everything matches nicely. That's much harder for me to do directly in RGB.

What do you think is a more intuitive model? I'm always happy to learn something new.

I think the wikipedia page is great--I've used the math on it to write conversion routines in the past.

If you just drop the HSL "saturation" to get a gray, you quite dramatically change the lightness in many cases.

More intuitive models:

Munsell Color System (related to rigorous experiment in the "Renotations" of the 40s. Many later models were attempts to model Munsell space with a closed formula instead of the lookup table that defines the Munsell System itself): http://en.wikipedia.org/wiki/Munsell_color_system

CIELAB, the de facto standard for many decades in colorimetry, used for example in many color management modules, defined in the mid-70s. (pretty weak article) http://en.wikipedia.org/wiki/Lab_color_space

IPT: I have no great links handy, but there's a mention here http://documentation.apple.com/en/color/usermanual/index.htm...

CIECAM02, the 2002 recommendation of the technical committee scientists studying this stuff, http://en.wikipedia.org/wiki/CIECAM02

Reply on news.ycombinator.com