Fwd: Convolution

[Boudewijn Rempt]

On Wed, 25 May 2005, Bart Coppens wrote:

> Well, there are still blue and red colored areas, but I thought those were
> fine? I'm all for moving to the new code, if, of course, we copy the 4
> matrices (1 for each channel), and use them, instead of hardcoding to matrix
> 0.

Are there actually any filters that currently use different kernels for
different color channels?

Boudewijn