M Y N D E X
Perception Research

We are currently researching perceptual contrast for assessing legibility and accessibility on graphically rich web pages and computer displays. Below are links to some of the early live experimental pages, evaluating the issue(s) in detail.


WEB APPS & STUDIES


PerceptEx   Δγ   The Web Perception Experiment

COMING SOON: a public study that you can be involved in! Currently under development, it will evaluate how people perceive content in real world situations. Stay-tuned here as we prepare to launch that project.


CVD Simulator    Color Vision Deficiency Simulation

NEW: We developed this new web app to simulate how someone with Color Vision Deficiency** sees the world. We accurately simulate Protanopia, Deuteranopia, and Tritanopia using the Brettel '99 &'97 models. We also present some experimental models, such as Blue Cone Monochromacity. www.myndex.com/CVD/

While the Brettel model is accurate for how certain types of CVD perceive the world, we also created an experimental model of how CVD specifically perceives sRGB monitors. Because tristimulus monitors only display three specific colors in various combinations, most of the colors that you see are metameric and non-spectral. So we also developed this sRGB CVD Simulator to simulate how someone with protanopia, deuteranopia, or tritanopia specifically sees sRGB monitors. www.myndex.com/CVD/sRGBCVD

**Color Vision Deficiency is sometimes referred to as "color blind,"
though that is not accurate except in some very rare cases.


Perceptual Contrast  ☼  Live Contrast Experiments 2019

These experiments relate to WCAG open issue number 695 and others on GitHub.#695 some pre-prints also on ResearchGate. For questions contact user Myndex on Github. (aka Color Science Researcher: Andrew Somers).

Below are links to some of the early live experimental pages, evaluating the issue(s) in some detail.

Please be advised: The experiment pages listed below are using fonts sized in REM and are thus NOT responsive. They are intended for desktop/laptop use, or use in our test app. Most require a monitor 1024px to 1200px wide to view.

CURRENT EXPERIMENTS

Most Recent Experiments First


NEW!   • CE16 SAPC-6 Polarity Test UPDATED! Nov 3 2019 (Polarity Experiments — Updated Nov 3, 2019 These are tests of the effects of normal and reverse polarity on contrast measurements. This is part of an assessment of the Luminance Contrast Module.
NEW!   • CE16 SAPC-6 Contrast Grid NEW! Nov 1 2019 (Dark text on Light BG) Perceptually Uniform grid of contrast patches, demonstrating SAPC-6 and also the cut offs for various font sizes and weights. This is part of an assessment of the Luminance Contrast Module. (Please note this page is NOT responsive, as it is intended for display on our research devices and uses REM to lock in certain sizes. As a result it may not work on your mobile device.)
NEW!   • Exp. CE16 — SAPC-6 Eval. NEW! Oct 27 2019 (Dark text on Light BG) Further perceptual uniformity Experiment. SAPC-6 is the latest advance in perceptually uniform contrast assessment. This is part of an assessment of the Luminance Contrast Module. This links to the short version with some discussion. Links to the complete study series are on that page.
   • Experiment CE14antialias (Oct 14,2019) Similar to "weight" below, only here the webkit antialiasing is turned ON, and you'll see the dramatic reduction in contrast for the thinner fonts, especially weight 100. In particular, look at Test Panel Four's first two series, and how the last two smaller weight 100 fonts practically disappear.
   • Experiment CE14weight (revised Oct 14,2019) How the stroke width (font weight) affects perceptual uniformity Experiment — font weights were grouped and resized for equivalent stroke widths. IMPORTANT DISCOVERY: The examples on this page demonstrate that when near threshold, stroke width is the primary "size" factor affecting CSF and contrast perception.
   • Experiment CE14size How Stimulus Size affects perceptual uniformity Experiment, also contains parts of CE14 for reference. The examples in this page demonstrate that stimulus size is a CRITICAL FACTOR in perceptual contrast, especially for impaired vision that is forced to function close to threshold.
   • Experiment CE15 — SAPC5 (INV) Evaluation (Light on Dark) Further perceptual uniformity Experiment. See CE14, this is the light text/dark BG version.
   • Experiment CE14 — SAPC5 Evaluation (Dark on Light) Further perceptual uniformity Experiment. CE14 and CE15 examine a new formula/algorithm for a perceptually uniform contrast assessment (SAPC-5). OF NOTE: This equation is showing overall promise in real-world ambient light, especially on mobile devices in daylight.
   • Experiment CE12 — SAPC3 Evaluation (Light on Dark v02) Perceptual uniformity Experiment. The sRGB values were adjusted for uniformity. Useful to note how polarity inversion affects perceptual contrast in terms of uniformity.
   • Experiment CE11 — SAPC3 Evaluation (Light on Dark). A straight inversion of the values used in CE11, see that when inverted the assessment does not indicate perceptual uniformity from bright to dark. See experiment CE12.
   • Experiment CE10 — SAPC3 Evaluation (Dark on Light). Sample DIVs with sRGB values that are approximately perceptually uniform with in each target range.
   • Experiment CE09 — Comparing several different contrast assessment equations.
   • (link off/available on request) Some initial equation evaluations and considerations. Early experiments solving the perceptual uniformity problem.
   • (link off/available on request) Experiments CEX and CEXI — A layout of the basic experiments but using the current WCAG math for a standard of comparison .
   • (link off/available on request) Initial Page describing the immediately apparent issues with the current standard & methods. Discussion of the genesis of this project, the discovery of the issues with the WCAG contrast standard math. NOTE: This page contains some of the initial reactions, but as with most research projects, many of the preliminary conclusions are superseded by more recent experimental results. Listed here for historical purposes only.


Brief Commentary on Contrast

and the WCAG Accessibility Standards.

June 2019:

Interestingly, the W3C/WCAG math is similar to a basic Weber contrast in terms of a perception curve when in a totally DARK environment (such as a bedroom at night). Of course the problems described are present in typical and high ambient lighting.

One of the curiosities with the WCAG math is the addition of a 0.05 "flare" component to both sides of the simple contrast equation: (L1 + 0.05) / (L2 + 0.05) This appears to do little except distort contrast by marginally pushing ratios slightly toward 1:1 (if you eliminate the L luminance terms, you get 0.05/0.05 = 1:1).

It is not a particularly useful way to boost contrast or limit low contrast. While it does help to slightly enhance contrast in dark color pairs, dark color pairs is where the overall WCAG still suffers most, especially in high ambient lighting conditions. Indeed this math generates some contrast figures that are ultimately illegible in real world conditions. (Paradoxically, the math also rejects some color pairs that are perfectly legible).

One of several alternatives, Hwang-Peli's Modified Weber adds the flare component to only one side, the denominator, as:

   (Llight - Ldark) / (Llight + 0.05) 

In our experiments in real world conditions (including mobile devices) we've found that an offset of 0.1 is more useful than 0.05 :

   (Llight - Ldark) / (Llight + 0.1) 

We discuss this modified Weber in this post in the GitHub discussion. This method provides a superior and more robust contrast assessment than the current WCAG methods.

Advances in Perception: Our ongoing research is working with a new set of equations & algorithms that are based on human visual perception. Beyond providing a reliable and robust assessment of contrast for accessibility guidelines for graphically rich web content, we anticipate further development of the concepts will lead to better environmental responsiveness for computer display devices in general.