Re: [ISSUE-2335] feConvolveMatrix bias property in SVG 1.1 SE

On 2010-12-02 06:31, Anthony Grasso wrote:
> ...
> So the suggested set of new equations is:
> aR = (aO (*) (kernel/divisor)) + bias
> cR = (cO (*) (kernel/divisor)) + bias*aR
>
> Please let me know if the above is your intended suggestion, particularly if my assumption for aR is correct. I would like to test what the output looks like with the new equations.
>...

This would indeed be my suggestion. I think there is more than one 
alternative that could make some sense, but the above is simple, very 
close to the original (it just adds the *aR at the end), and makes the 
output more useful, in my opinion.

BTW, when you're going to experiment, also try preserving the alpha 
channel. (And I think Inkscape still implements the above, so you don't 
necessarily have to code anything yourself to try it out.)

Received on Thursday, 2 December 2010 09:52:24 UTC