Web Camera processing

Hi,

If I read a 1.3mp webcam image and collect statistics, cut out the
centre 640x480 or scale the image and store that with an algorithm
applied to each new pixel, then stack the last 16 images and put put
those to the S-video port how many frames a second could I capture
from the input Webcam stream, ESTIMATED.
I assume I can read the last Webcam buffer when I am ready discarding
missed ones, and that I can re-fresh the S-video buffer when I am
able, maybe having to refresh the last accumulated image till the
next
one is ready.
I would be using the GUI on the monitor for controlling the applied
algorithm.

Cheers. Andrew.