Neuromorphic (event-based) vision sensors are expected to advance the frontiers of high-speed, high-dynamic-range and low-power vision-based robotics in the upcoming years. Event cameras are becoming more and more affordable, and are now widely available in robotics labs where their utility has been confirmed via experiments. In spite of these empirical successes, skepticism persists regarding the superiority of event cameras to conventional frame-based cameras in light of the concurrent developments of low-cost, high-performance CMOS technologies. We propose an information-theoretic comparison of event-based and frame-based vision sampling schemes, which we believe will formally prove (or disprove) the advantages of different schemes.
