From: Rik Cabanier <cabanier@gmail.com>

Date: Wed, 25 Jul 2012 20:59:19 -0700

Message-ID: <CAGN7qDB_wgVUgs2puK4WkjcXzaN6Zg2Red3K8Xnm0+7mqtSe9Q@mail.gmail.com>

To: Calculemus <calculemus1988@gmail.com>

Cc: steve@fenestra.com, www-svg@w3.org

Date: Wed, 25 Jul 2012 20:59:19 -0700

Message-ID: <CAGN7qDB_wgVUgs2puK4WkjcXzaN6Zg2Red3K8Xnm0+7mqtSe9Q@mail.gmail.com>

To: Calculemus <calculemus1988@gmail.com>

Cc: steve@fenestra.com, www-svg@w3.org

That looks correct. Try to debug your code with solid color bitmaps (the top one should be 75% alpha) and see if your output matches the formulas and photoshop's/gimp's rendering. Rik On Wed, Jul 25, 2012 at 5:59 PM, Calculemus <calculemus1988@gmail.com>wrote: > Hmm, actually I do take care of that. > > These are the formulas I use: > > Notation: > > a - premultiplied color > alpha_a - alpha for a > same notation for b > > color = a*(1-alpha_b) + b*(1-alpha_a) + alpha_a*alpha_b*Blend(a/alpha_a, > b/alpha_b) > alpha = alpha_a + alpha_b*(1-alpha_a) > > where Blend is replaced with the proper blend formula as needed. > > I don't divide 'color' with 'alpha' because we use pre-multiplied values. > But I do provide Blend with non-premultiplied values, that is why I divide > there. > > >Received on Thursday, 26 July 2012 03:59:46 UTC

*
This archive was generated by hypermail 2.3.1
: Wednesday, 7 January 2015 15:29:51 UTC
*