W3C home > Mailing lists > Public > public-apa@w3.org > July 2016

Re: Color Contrast and Transparency | WCAG 2.1

From: Detlev Fischer <detlev.fischer@testkreis.de>
Date: Fri, 29 Jul 2016 11:02:01 +0200 (CEST)
To: w3c-wai-gl@w3.org, public-apa@w3.org, acampbell@nomensa.com
Message-Id: <20160729090201.BAE3F6820202@dd24924.kasserver.com>
In testing, we often encounter text over photos which have some parts of text with sufficient contrast, others not so.
The problem may be minor (just some small patches with insufficient contrast). If the background is not largely uniform / blurred but irregular, chosing the reference point for the colour picker gets difficult. In many cases, picking the worst bit of background (as David suggests) seems subjectively unfair if we are dealing with a small and isolated patch. Its main benefit is that it is rule-like ("always go for the point with the weakest contrast even if it is just one pixel wide").

Theoretically one could create a mix of bg colours (use Gaussian blur) to get the 'average' bg colour but that would not account for the perceived difficulties for reading on vivid backgrounds where the averaged value would pass but the text is nevertheless hard to read.

In sum: measuring contrast of text on photo backgrounds involves subjective judgment. Tools help, but whether you pass or fail an edge case around 4,5:1 will be down to your handling of the colour picker...

Detlev Fischer
testkreis c/o feld.wald.wiese
Thedestr. 2, 22767 Hamburg

Mobil +49 (0)157 57 57 57 45
Fax +49 (0)40 439 10 68-5

Beratung, Tests und Schulungen für barrierefreie Websites

Alastair Campbell schrieb am 29.07.2016 10:39:

> JF: “the impact of Alpha Transparency of colors and the impact on color
> contrast and accessibility?”
> I guess this is mostly a matter of how it is measured?
> So if designer/developer specifies a foreground and/or background colour with
> some transparency, you shouldn’t use the CSS defined colour to test contrast
> without accounting for the transparency. (I know transparency is generally
> defined with the colour, but that doesn’t mean a testing tool accounts for
> it.)
> If the transparent colour of a box (the background for text) is on top of a
> white background it will be lighter than on a black background, which needs to
> be accounted for.
> However, if you use a tool with an eye-dropper style mechanism on the resulting
> colours, that should be accurate. Unless I’m missing something?
> The contrast adjuster function in the CSS-color-4 spec looks cool (if it’s
> implemented?), it references WCAG2 1.4.3 so should line up. 
> Perhaps we need to make G18 more specific about testing the result rather than
> the colour definition?
> And perhaps we could create a test page that would pass if transparency is not
> accounted for, but should fail, and check that each of the tools listed in G18
> accounts for the transparency? (Perhaps the browser based ones don’t?)
> Cheers,
> -Alastair
Received on Friday, 29 July 2016 09:02:31 UTC

This archive was generated by hypermail 2.4.0 : Thursday, 24 March 2022 20:23:01 UTC