Color Vision Deficiency Test

Normal color vision samples the spectrum through three cone photopigments — L, M and S — and a color is nothing more than the triple of their responses. Convert a screen color to that space and you can do something specific: move it along exactly one of the three axes and leave the other two untouched.

Two colors that differ only in their L response are, to someone with no L cone, the same color. Not similar — identical, because the signal that told them apart was never received. Those pairs lie on that observer’s protan confusion line. The same construction on the M axis gives deutan pairs and on the S axis tritan pairs.

sRGB  →  linear RGB  →  LMS          (Hunt–Pointer–Estevez)

figure      = LMS with one axis × (1 + δ/2)
background  = LMS with the same axis × (1 − δ/2)

δ large  → coarse plate, only a strong deficiency misses it
δ small  → fine plate, mild anomalous trichromacy misses it too

The dichromat simulation used to verify each plate is the standard one from Viénot, Brettel & Mollon (1999): it discards the missing cone signal and reconstructs it from the other two. Because that reconstruction is a linear map, a pair differing on only one axis maps to the same output — which is the property the test suite in this repository asserts on every plate before it ships.

Worth knowing what this page is an approximation of, and where it stops.

1. 01

   Screening plates

   Pseudoisochromatic plates, calibrated and printed, viewed under a standard illuminant. Fast, and they detect the presence of a red-green deficiency reliably. This page is a browser approximation of this stage and only this stage.

2. 02

   Arrangement test

   The Farnsworth D-15 or Farnsworth–Munsell 100 Hue: the patient orders colored caps by hue. The pattern of errors falls along the confusion axis, which classifies the type and gives a rough severity.

3. 03

   Anomaloscope

   The definitive instrument. The subject matches a mix of red and green light to a yellow reference. The mixture ratio they accept identifies the type exactly and quantifies the shift. No screen-based test can replicate this, because it depends on calibrated monochromatic light.

4. 04

   Occupational testing

   Where a license depends on it — aviation, maritime, rail — a lantern test such as the Holmes-Wright or CAD test is used, because the practical question is whether signal colors can be identified, not what the underlying genetics are.

• Control plates missed — the result is reported as inconclusive rather than as a deficiency. Control plates differ only in lightness, so color vision plays no part in reading them. Missing one points at brightness, a color filter, or a mistyped answer.
• One fine plate missed on one axis — within normal variation. Plenty of people with ordinary color vision miss a subtle plate on an unprofiled laptop screen, which is why the scoring does not flag it.
• Fine and moderate plates missed on one axis — the pattern consistent with anomalous trichromacy on that axis. Common on the deutan axis.
• All three plates missed on one axis — the pattern consistent with a strong deficiency or dichromacy on that axis. Worth a real examination.
• Protan and deutan both low — expected. The two red-green types share most of their confusions, and reliably separating them is precisely what an anomaloscope is for.

If you arrived here looking for the plain-language version — prevalence, everyday impact, what to do next — that is the color blind test. To see what a given image looks like to a protan, deutan or tritan viewer, the color blindness simulator at hextorgb.io applies the same Viénot transform used here, in reverse.