- From: Simon Sapin <simon.sapin@kozea.fr>
- Date: Mon, 29 Oct 2012 07:58:59 +0100
- To: Lea Verou <lea@w3.org>
- CC: w3c-wai-ig@w3.org, Chris Lilley <chris@w3.org>, Shadi Abou-Zahra <shadi@w3.org>
Le 29/10/2012 00:14, Simon Sapin a écrit : > We do want something "close" to the text color, but in luminance space, > more or less. I went a bit too fast and did not prove this. > You already proved the bounds of the luminance for the > background blended on its backdrop: they are reached when the backdrop > is black or white. We only need to the background luminance closest to > the text’s, but within these bounds. The contrast ratio of two luminance values is defined as: C(L_a, L_b) = (L1 + 0.05) / (L2 + 0.05) given L1 = max(L_a, L_b) L1 = min(L_a, L_b) Now let’s consider C(L) = C(L, L_text) the contrast between a variable blended background lumimance L and a fixed text luminance L_text. We are trying to minimize C(L) * At L = L_text, C(L) = 1 with both "legs" (C is continuous.) * For L > L_text, C(L) is affine with positive coefficient and is therefore strictly increasing. * For L < L_text, C(L) is the inverse of a strictly increasing affine, and is therefore strictly decreasing. Therefore, L = L_text is the absolute minimum of C(L) and a smaller abs(L - L_text) ("distance" in luminance space) implies a smaller C(L) (smaller contrast ratio) > If the text luminance*is* within these bounds, the minimal contrast is > 1:1. Otherwise it is either higher than both (minimal contrast is > obtained on a white backdrop) or lower then both (minimal contrast is > obtained on a black backdrop) -- Simon Sapin
Received on Monday, 29 October 2012 06:59:40 UTC