From: Rik Cabanier <cabanier@gmail.com>

Date: Sat, 28 Jul 2012 12:52:45 -0700

Message-ID: <CAGN7qDCOr0VTLGKxB1nPyEKm3-Rv1iCZKc6GO0wf6Gbz+DM-Qw@mail.gmail.com>

To: Calculemus <calculemus1988@gmail.com>

Cc: Dirk Schulze <dschulze@adobe.com>, "steve@fenestra.com" <steve@fenestra.com>, "www-svg@w3.org" <www-svg@w3.org>

Date: Sat, 28 Jul 2012 12:52:45 -0700

Message-ID: <CAGN7qDCOr0VTLGKxB1nPyEKm3-Rv1iCZKc6GO0wf6Gbz+DM-Qw@mail.gmail.com>

To: Calculemus <calculemus1988@gmail.com>

Cc: Dirk Schulze <dschulze@adobe.com>, "steve@fenestra.com" <steve@fenestra.com>, "www-svg@w3.org" <www-svg@w3.org>

Let's do the math! For an example, take a pixel that is RGB(0,1,0,.5) that is on top of RGB(1,0,0,1) with a multiply blend mode. (This is how photoshop/gimp are treating your example) First you alpha composite the pixel with its backdrop, so: co = Cs x ás + Cb x áb x (1 - ás) = RGB(0,1,0) x .5 + RGB(1,0,0) x 1 x (1-.5) = RGB(.5,.5,0) áo = ás + áb x (1 - ás) = .5 + 1 x (1 - .5) = 1 Now you blend this: B(Cb, Cs) = Cb x Cs = RGB(.5,.5,0) x RGB(1,0,0) = RGB(.5,0,0) Feed this into the full compositing formula: c0 = ás x (1 - áb) x Cs + ás x áb x B(Cb, Cs) + (1 - ás) x áb x Cb = 1 x (1-1) x RGB(.5,0,0) + 1 x 1 x RGB(.5,0,0) + (1-1) x 1 x RGB(1,0,0) = 0 + RGB(.5,0,0) + 0 = RGB(.5,0,0) Now take a pixel that is RGB(0,1,0,1) that is on top of RGB(1,0,0,1) with a multiply blend mode and alpha of .5. (This is how you are doing the math) First you blend this: B(Cb, Cs) = Cb x Cs = RGB(0,1,0) x RGB(1,0,0) = RGB(0,0,0) Feed this into the full compositing formula: c0 = ás x ((1 - áb) x Cs + áb x B(Cb, Cs)) + (1 - ás) x áb x Cb = .5 x (1-1) x RGB(0,1,0) + 1 x RGB(0,0,0) + (1-.5) x 1 x RGB(1,0,0) = RGB(0,.5,0) + RGB(0,0,0) + RGB(.5,0,0) = RGB(.5,.5,0) Notice that in the first example, you have to composite twice. Rik On Fri, Jul 27, 2012 at 4:44 PM, Calculemus <calculemus1988@gmail.com>wrote: > Rik, > > 1. "Where is the alpha coming from? Is it already in the image that you're > applying blending to?" > > I created the images A and B in Photoshop, A has opacity > 75% and B has 100%, so the alpha is coming > from the images yes. > > 2. "If so, you need to alpha blend with the background first and then > blend." > > You mean alpha blend the top image A with bottom/background image B? > And then blend with...? I am confused, so I am not supposed to use that > formula? > > Thanks >Received on Saturday, 28 July 2012 19:53:13 GMT

*
This archive was generated by hypermail 2.3.1
: Friday, 8 March 2013 15:54:51 GMT
*