Multi-imaging and Bayesian estimation for photon counting with EMCCDs

A multi-imaging strategy is proposed and experimentally tested to improve the accuracy of photon counting with an electron multiplying CCD (EMCCD), by taking into account the random nature of its on-chip gain and the possibility of multiple photodetection events on one pixel. This strategy is based on Bayesian estimation on each image, with a priori information given by the sum of the images. The method works even for images with large dynamic range, with more improvement in the low light level areas. In these areas, two thirds of the variance added by the EMCCD in a conventional imaging mode are removed, making the physical photon noise predominant in the detected image.