How to choose the right interface for an embedded vision system?