Hacker News new | ask | show | jobs
by hansvm 695 days ago
I'm not sure about straight-up glasses, but there are a few effects you can exploit to enhance hue distinction in the normally afflicted regime (necessarily always at the cost of some other sort of distortion).

The simplest example is just (for red-green colorblindness) red-tinted or green-tinted glasses. You lose some dynamic range and add a distortion across the spectrum, but you gain brightness as a channel whereby you can distinguish many reds from many greens. I used to use strongly orange-tinted glasses to help spot orange buoys in the ocean, which otherwise wouldn't stand out very well against the greenish notes in the water. They'd appear quite bright against a dark background.

A usually less disruptive and less effective strategy is something like what enchroma does. Real objects aren't just red or green. They reflect a variety of rays across the spectrum. If you add a notch filter to remove the light between the two affected cones, green objects will activate the green cones nearly the same as before and the red cones less than before (and vice versa for red objects). That doesn't matter for pure colors like a red or green laser, and it doesn't allow you to see new colors, but it makes most red things a little more red and most green things a little more green, slightly improving hue discrimination (also making some objects nearly black if they hit that notch filter exclusively).

And so on and so forth. Normal lenses won't have any photon upconversion or downconversion or anything, so your only available mechanism is altering the transmission spectrum, blocking some wavelengths more than others, perhaps differently in different eyes or parts of your field of vision. You can give up some other information to add a little resolution to that hue distinction information channel.

With computers in the mix, you can tune that a little better by also shifting hues. Doing so "correctly" is a little hard though. You'd like to do something like just move the problematic wavelengths somewhere else (ideally bijectively, squishing other wavelengths to make room for the displaced ones), but your computer monitor doesn't output a whole spectrum of wavelengths; it outputs mixes of three (three distributions really, but we don't really care). Red and green are two of them, and those are the very things you'd like to distinguish better, so your only real option (realness being measured in ability to help you strongly distinguish red and green) is to add or remove some extra blue (e.g., adding a lot when red dominates green or a little at the other extreme) or to alter the brightness in response to red/green discrepancies. Most OS's want to keep brightness invariant in their colorblindness settings, so the only option is messing with the blue channel. Doing so removes your ability to distinguish some colors relying on the blue channel and adds the ability to tell red from green, with varying degrees of efficacy depending on who coded it and how many details they accounted for.

You could, in theory, have something like a VR (AR? You're augmenting the existing data, but the tech capable of doing that is closer to a VR headset) experience to do a transformation like that on real-world visual data. Maybe simpler when climbing, AR could add stripes or some other distinguishing feature to handholds of a certain color.

Anywho, maybe try sunglasses strongly tinted as the color of the handholds you'd like to spot.

1 comments

Thanks for the write-up. As far as I can tell, what you are describing in your first three paragraphs is exactly what I was trying to describe? Block specific wavelengths. (EDIT: sorry, I was mixing things up. My comment that I was referring to is https://news.ycombinator.com/item?id=41066579 but yours wasn't in reply to it.)

I don't think wearing any goggles with screens in them is realistic for climbing at the moment.

Yes, adding more distinguishing factors to the climbing holds works. A laser pointer's dot can be seen as a very temporary distinguishing mark.