During the last decade, the CMOS camera has developed rapidly and quickly became a commercial success. A number of research groups are developing quite different CMOS vision sensors: bio-inspired, so called 'retinomorphic' or 'neuromorphic' cameras. They operate quite differently and offer various new functionalities, however, they cannot yet match commercially developed products in accuracy and resolution. In 2005, an AER retinomorphic camera was produced at the University of Oslo which, though it was a successful project, suffered from a lot of salt-and-pepper noise. Due to the amount of logic they contain, retinomorphic pixels are already quite large compared to commercial photo pixels. Therefore, the goal of this project was to reduce the noise of a retinomorphic pixel without increasing its size. An improved pixel design was then used to make a camera on an integrated circuit.
A secondary project is included in this thesis as well. Until now, all retinomorphic AER pixels have been monochromatic and this project investigates the possibility of making a dichromatic retinomorphic AER pixel. Stacked photodiodes exploit the fact that photons of longer wavelengths are likely to penetrate deeper than photons of shorter wavelengths, and implementing this solution into a retinomorphic AER pixel was the goal of this project. A test pixel was realised on an integrated circuit.