Blending-Mode Label Coloring
Last winter, I was experimenting with applying blending modes to label text, as a way of annotating a map without a heavy-handed solid fill or double stroke. I wanted the labels to recede when looking at the map as a whole, but then appear if the eye stopped for a closer look.
Labels that Know Their Place
I have mixed feelings about labels -and I almost never use them in my maps. But when I do, I wrestle with how to design them so they A) don’t steal the show, and B) are visible over an impossibly varied background.
So by painting labels with the pixel values of the map underneath, with some hue/saturation/lightness levers pulled, I know the label will always have a boosted appearance, relative the the busy context directly below. The effect can be strong or weak (I of course liked the weaker effect).
Elevation and Color
To test blending-mode labeling, I used Digital Elevation models (I like these things). They provide an inherently varied background, but, when styled minimally (just changing the color gradient) they have a ghostly appearance that seemed to match the vapory look of the label experiment. So I got carried away in the testing and just made a bunch of blend-labeled, colorized, elevation maps.
Click to embiggen…
Cheers, John
Adventures In Mapping 
I noticed, that some labels seem to disappear completely when zoomed out.
Could this be extended into some sort of ghost labelling that is practically invisible at certain distances?
Just like printed CMYK halftone screen pattern become invisible.
So having high frequency darker Text-(Outline) plus brighter halo (or vice versa) cancel out to the average of the surrounding hue and lightness at lower frequencies?
This should be doable by somehow using wavelet (also chirplets, curvelets?)-like transformations (all transformation, where the sum/integral of differences is zero) on the label with the background color as neutral level, that is optically neutral at longer viewing distances (lower frequencies), but looks like labels on closer distances.
When choosing different bandwidths for the transformation, smaller bandwidths should disappear faster with decreasing viewing distance.
http://www.curvelet.org/gallery.html
This could also be applied to all kinds of charts.
After all, this is just a wild thought on a caffeine-heigh, but I think its worth sharing.
LikeLiked by 1 person