Hacker News Reader

Comment by roel_v | Hacker News Reader

What does 'telling a color apart' even mean? I did a mini project some time ago, where I basically did some custom color quantization (it was to reduce an image into a configurable amount of discrete colors, to make a pixel quilt with a manageable amount of distinct fabric colors). So I used CieLAB colors to calculate the 'visual distance' between the available colors and then recursively merged colors until I got to the desired amount. But even a simple algorithm like that, it turned out, could be implemented in several way; each of which gave different, yet equally 'valid' results.

My point being - it's very hard to give a simple, unambiguous meaning to seemingly simple concepts like 'what is a color', 'when are colors the same' and 'are these two colors more similar than those other two'. There is a lot of abstraction between the physics of light and representations of color e.g. on a screen, or in fabric, or when printing.

View on HN

> What does 'telling a color apart' even mean?

Being able to distinguish colors on a test such as https://xritephoto.com/cool-tools

Although I got it right in the end, I found the second row to be insanely hard.

But somehow, on all rows, the colors just seemed to snap into place for me, is that a normal feeling?

I think in the case of telling two colors apart, it's can be as simple as presenting them side by side, and seeing if someone can distinguish a difference.

Not sure that is so simple unfortunately.

First, because if I recursively apply your argument, then no color would have a name. Starting from the very 1st lowest color frequency and the 2nd right next to it, no-one could differentiate them. Apply the same logic between the 2nd and the 3rd, the 3rd and the 4th, etc.

So you cannot apply this argument to "any" color pair. You would have to define a set of starting colors, and always do the comparison against them. Now you enter the problem of which starting colors to starts, and the fact that using any pre-existing color names would most likely not be evenly distributed on the color spectrum.

Second, it wouldn't work because color recognition is not a transitive process.

- If you show me two very close red variations side by side, I may not be able to tell the difference.

- If you space the variation, I may be able to tell that there are 2 different reds, but wouldn't be able to tell which one is darker/lighter accurately.

- If instead of showing me the two close reds side by side, you show them to me paired with an other color, I may be able to tell the difference. For instance, show me a board with half red1 + yellow, and an other board with red2 + yellow. I may spot than 1 of the red + yellow is matching that good compared to the other one.

You can get out of that argument by saying that colour names apply in different amounts to different hex-colours.

So #FF0000 is red. But is #FA0101 is also red, just slightly less so.

That's interesting. So what you have is some sort of similarity metric that is colour based. You could define it as being 1 for an exact match, and 0 for not matching, so cyan, etc. would be 0 red, while orange may be slightly red (e.g. 0.1), and pink may be slightly more red (e.g. 0.25). This would be a 3D function/surface.

This would allow you to model classes of colours, like purples. Shades of a colour could then be any similar colour in the range 0 to 0.5 (or 0.25).

Yes, this is what Delta E in the CIELAB color space is meant for. But in RGB color space this concept doesn't make sense; well, only if you would accept a non linear similarity metric - but then what is the metric for, if you need to apply a curve to it to normalize it? But still, this is different from modeling 'shades' of a color (if we say 'shades' are varying brightness in RGB space using the 0.21/0.72/0.7 brightness conversion - but again, I argue that 'shades' is an ambiguous concept).

My point is, again, that colors are a much more difficult concept than some in this discussion are making it out to be, and 'distance between colors' is more complex still.

Yes, agreed.

Yep, in fact, that’s exactly what those hex codes are: a specification for how much red, how much green, and how much blue is in the color (though our vision and color perception may not map linearly to that specification).

Sort-of. My aim was to suggest looking at natural clusters of similar shades.

You can map those clusters into RGB space. Or other colour spaces.

Technically you are correct, but I don't think that's what the colour-namers are going for.

That restricts you to 8 bits per color channel, and more importantly, to the RGB color space which is non linear. Comparing colors in RGB color space is nonsense in the first place ('comparing' here is 'judge by human perception'). 'Comparing' colors is just not that easy; in fact, I'd argue that without (much) more definition and clarification, it's impossible to do, because it can mean different things depending on the purpose.

I only used RGB and 8 bits per channel in the example.

My argument was more general.

Reply on news.ycombinator.com