2 ms·
I agree that this blog post takes a sort of silly tone, and also that many artists are very sophisticated in their understanding and use of color. He goes wron
by jacobolus 14y ago
I agree that this blog post takes a sort of silly tone, and also that many artists are very sophisticated in their understanding and use of color.
He goes wrong in a few places, for instance in using quite confusing coloration of his diagrams, and he really shouldn’t put the labels R, G, B on cone responses: these are too evocative and are likely to confuse as much as they enlighten.
But to be fair, basic paint mixing models in artists’ heads often aren’t very good, and many artists would really benefit from studying human color vision more seriously. Explaining how vision works to visual artists, especially amateurs, is one of the more rewarding things I ever do: you can see these lightbulbs going off in their heads as a bunch of stuff that they know “intuitively” and experientially is given proper names and ordered into a comprehensible model.
> As for CMYK and RGB color spaces, these are simple but quite effective tools for dealing with color in a purely mechanical way.
I think these spaces have no place being taught to non-specialists. They are not intuitive to humans, being based on particular technology rather than on human vision. Any user interface which exposes these color spaces to non-experts is letting them down. Just as bad is using HSL or HSV or some similar trivial derivative space. Image editing software, in particular, is dramatically harder to use than it would be if the dimensions were at all relevant to vision.
> If you really need a color model more closely tied to how human visual physiology works, LAB was designed as a good approximation of that. Specialists sometimes use other, often proprietary systems.
Specialists looking for models related to vision mostly use open, publicly developed and specified models. At root, most of modern colorimetry is based on the CIE system, which has served very well since the 1930s. Recent fairly effective models that would be useful for software include CIECAM02 and IPT, which behave better in some ways than CIELAB, but are a bit more computationally expensive.
> If you need this level of knowledge, you got some book larnin' to do. No blog posts or comments on HN is going to help you too much.
That’s reasonable. Let’s start with some pointers to good sources!
For this audience, Maureen Stone’s SIGGRAPH notes from 2001 are a nice first start, with lots of pointers to other good resrouces: http://graphics.stanford.edu/courses/cs448b-02-spring/04cdrom.pdf http://graphics.stanford.edu/courses/cs448b-02-spring/04cdro...
For anyone who wants to really dive deep, I recommend http://www.handprint.com/LS/CVS/color.html http://www.handprint.com/LS/CVS/color.html (edit: Actually, I see that the blogger does link there. So that’s nice.)
Also, anyone looking for pointers to books, feel free to email me (address in profile), and I can point you in some direction based on your specific interests.
- commieneko 14y agoArtist's color mixing models aren't simple mechanical ratios. They use emergent properties of chemicals that just happen to be a particularly useful color. In fact, artist try to do as little color mixing as possible. Mixed pigments of the types artists use do work like the inks used in printing and the dyes used in RGB displays. Color wheels are guides and approximations. Learning a color theory that is more accurate is counter productive because you can't find pigments that behave that way. And if you ever get a chance to see a painting in the flesh and compare it to even a very good reproduction, you will see that the simple color model primaries used in printing and on screen reproduction are only loose approximations of the real colors. Now what an artist does need to know is how color works perceptually. And the models used for this are pretty good. And remember the intended result is seldom "photo realism"; not that photos are particularly realistic. But that's another discussion altogether. If you are talking about amateur artists, or beginners, then, sure, lecture away, but if you talk to someone who's been successfully painting for 40 years, my advice is to listen to him/her. They may or may not know much about the physiology of human vision, but don't be surprised if they do, but they certainly know the ins and outs of making you think you see what they want you to see. The nice thing about standards are there are so many of them. I've spent more time than I cared to writing software to convert and compare between systems. They are a few lingua franca systems, but if you are doing serious, critical color work, your special cases are going to be the bulk of your job. Which only makes sense if you think about it. The sources I used are going to be about 30 years out of date, and the chemistries and optical systems no longer relevant. Doing mechanical color reproduction is a specialty and few people need to know more than the basic RGB/CMY model and whatever specific systems that they need for their domain of activity. These days these models are very good and well tuned for their domains. Unless it's a brand new technology (Or NTSC; don't get me started!) This usually means working with a specific palette and color gamut. And while a lot of people in the various industries feel that they are dealing with the real color models, and the rest mere approximations, there is no such thing. In the end it all breaks down and you are left with making specific measurements with specific types of equipment for use with specific use cases. And then you calibrate. Color theory is very much a case of the map not being the territory.
- jacobolus 14y ago