jheriko:
i though about doing dynamic convolution a bit ago and the only practical way i reckon it can be done is using 2d fourier transforms, i was going to do this myself but i'm not sure if i'll ever get round it to it now so feel free to have a go.
basically if you take the (2d) fourier transform of a convolution kernel and multiply it point by point by the fourier transform of some image, then do the inverse fourier transforfmation, you are left with the image convolved by the kernel. thus you could use anything you can generate with a superscope dynmov etc as your kernel, thus it could be as dynamic as you like.
your best bet is to use the fftw libraries from
www.fftw.org (i think)
i dunno how much you know about fourier transforms, i seem to remember you being fairly "up" on your maths. i think the fftw site probably explains them if you don't know much.
hope this helps.