Quartz Extreme/Core Image: applicable in robot vision?

I have been doing some experimentation with Quartz Compositor and Core Image, and it seemed to me like a big, big promise to the field of robotic vision. I am working on a Macbook 2GHz with 950GMA integrated graphics, not as good as the current Mac Pro, Macbook Pro and iMac offerings, but I am seeing very high levels of performance.

A Windows PC does not yet use the graphics card to accelerate video processing. While I know that some robot vision is developed in Matlab, I know that it takes the whole 100% of CPU time to extract relevant vision clues for robot control. An OSX machine may be able to take these tasks and only consume a fraction of CPU power (while the GPU heats up).

Anybody here working on robot vision? I'd like to grab some algorithms and test whether they would be significantly accelerated with Core Image? Any Core Image experts here?