Optimising number resolving photo-detectors using classical post-processing

Many present day quantum optics experiments, particularly in optical quantum information processing, rely on number-resolving photo-detection as a basic building block. In this paper we demonstrate that a simple classical optimisation technique can sometimes be employed to post-process the detector signature and improve the confidence of the measurement outcome in the presence of photon-number errors such as loss or dark-counts. While the regime in which this technique is applicable is rather restrictive, and will likely not be very useful for the large-scale quantum information processing applications of the future, the ideas presented might be employed in some present-day experiments where photo-detectors are typically very poor.