1

I'm trying to choose a discrete set of 10+ "cold" to "hot" colors to be used by translucent markers/icons displayed over a continuous tone grey scale image where each step is clearly different it's the neighbors. (i.e. not just some smooth gradient.)

Being perceived as "cold" to "hot" is not as important as each level being distinct from the next.

Some of the more naive choices have at least some colors that differ mostly in "brightness" (?). This can cause confusion when the markers from these colors blend with the underlying grey values.

Is there a known palette (or method to generate a palette) to use where no matter how the colors blend with the grey values, each bin remains clearly distinguishable from all others?

For an example of usage:

  • Base image is a black/white image representing the relative landscape elevation for an area of rough and knotty terrain.
  • Translucent marker icons would be indicating a spot measurement of temperature or gas density.
  • Markers are colored according to bins of value over the actual range. e.g. If the captured values are 10 to 100 degrees across all markers, the colors would be: 10-19 degrees is blue, 20-29 degrees is green, [...], 100+ degrees is red)
  • Markers are not opaque, since the underlying image features needs to be visible through them.

The question is how to select a good ~10 colors for the translucent markers such that they remain distinguishable no matter the underlying grey value.

Seems like the sort of thing that would be known in the data visualization field, but my Google skills are failing me here.

Thank you!

  • I can't visualize your requirement. You want 10 different semi-transparent coloured icons to show up clearly on something like a black & white photo? – Stan Jun 7 '16 at 1:19
  • These 10 different colours must be in the blue to red range of hues? – Stan Jun 7 '16 at 1:22
  • Can you give more information such as the use of the visual device. Is it to mark location, indicate an object, help organize the image in some way? Will all of the things be used at once? Will they overlap or align to obscure each other? Do they have to be the same size and/or shape and/or orientation? – Stan Jun 7 '16 at 1:35
  • @Stan I updated question. Hope this clarifies things. – Wafflecode Jun 8 '16 at 15:01
  • I had to do oil-well patch maps for a large Canadian company. Visualizations were not a cakewalk as depth information and drilling patterns were to be presented on one map. Thankfully, there was no surface information to add. In my experience, 6 is the maximum number of hues that can be unambiguously recognized. – Stan Jun 8 '16 at 17:52
0

The problem of using "lightness differences" of semi-transparent coloured icons/markers on a background of varying lightness is one will interfere with the other at some point.

No problem with up to 6 different colours. After that, you must make concessions which adds to the ambiguity between one colour and another.

You have a better chance of success if you can use some other attribute in addition to hue variation of your icons/markers. Anything you can add to the icon/marker colour will help define the variable: Size, Shape, Texture.

  • Yeah, the more non-color cues I can use the better. But I am limited there. You say: "No problem with up to 6 different colours". This gets to the heart of my question. Is there a simple axis in some color space I can use which is independent of grey scale blending. I was thinking Hue from HSB, but I don't know if that's really true. Then we'd just be left with ensuring enough separation exits between bins. Which may or may not be doable per your comment. – Wafflecode Jun 10 '16 at 17:16
  • @Wafflecode Six is a number that was used in discussions about UI research and perception issues with colour discrimination. Don't forget that 10% of males with Red-Green and 3% Blue-Yellow have to be accounted for in your display. Females have the same proportions but lesser %age discrimination to hue variation. Sorry to kill your day with that nugget. – Stan Jun 10 '16 at 18:45
  • @Wafflecode The take-away is that you could find 10 different samples; but, if your users can't see more than 7 unambiguously, you're not solving the perceptual "problem." – Stan Jun 10 '16 at 18:51
  • Interesting. If 6 is the useful upper limit, maybe we need to think about some kinder of coarser binning method. Frankly, that might be fine too. But even for those 6, can you point me to some research on the subject? Is there a specific known-optimal set, or is it a family the meets certain criteria? – Wafflecode Jun 14 '16 at 16:41
  • @Wafflecode This is from my notes on perception. This isn't a hard number but a guide. Testing will be more conclusive for your particular application. Note that there are saturation limits for almost any kind of stimulation. Discrimination within that point will vary with the individual and even with an individual (fatique, mood, motivation, ambient conditions, etc.) – Stan Jun 14 '16 at 22:39
0

To accomplish the overlay effect and keep colors saturated but transparent, change the marker's/icon's opacity from Normal to Multiply in any Adobe program.

To find a good set of colors to use, you can use the Color Guide in Illustrator to see a bunch of color options that go with a color you've already selected (ex: select red in the swatches, then click the Color Guide, then go through each dropdown to see a variety of shades, tints, complimentary colors, etc. and you can be sure that the colors will all go together but have enough contrast between them). Or you can try kuler.com and browse different color schemes or create your own.

0

If you can implement a two-part tag with an underlayer which is rgb(128,128,128) @ 50% opacity and set to an "exclusion blend mode", and then the colored tag @ 50% opacity, you will wind up with a tag that maintains its color identity on a very wide range of grey values.

This will give you some leeway for picking tag colors.

It will modify the contrast a little of the underlying image but it will be minimal and will be masked by the tag color.

The top square is without the exclusion layer.

enter image description here

  • That's a great sample. Thank you. In my case the markers are smaller, but you got the idea right. That's an interesting backup plan. Rather than ensuring identity at the palette level, we can do it at the blend level. It would mask some detail in the underlying image, but that may be a necessary compromise. – Wafflecode Jun 10 '16 at 17:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.