Reducing Collision Noise In Asynchronous Vision Chips

This paper presents a Monte Carlo evaluation of the quality of information extraction from event based from intelligent vision chips. Such systems have many advantages over traditional raster scanning. However, we show how collisions in the reading out of events, leads to major degradation of image quality. These effects have been modeled in terms of scalability and energy consumption. We propose a theoretical system level scheme to reduce these issues and open up the possibility of using intelligent vision chips at mega pixel resolutions.