Adaptive motion detection algorithms

We need one or two (15-30 hp) students for designing and implementing a robust motion detection algorithm. The algorithm should be able to adapt to the current environment in terms of ordinary background noise.

Ideally it would be able to take into account things like:
* activity on the street outside the window
* house pets
* weather

Secondly we also want to use image recognition to pick out the most relevant image from a series of pictures according to certain criteria (like contains a face looking into the camera).

We’re looking for energy efficient solutions that can be implemented in either Java or C.