Hot pixel classification of single-photon avalanche diode detector arrays using a log-normal statistical distribution

CMOS single-photon avalanche diode (SPAD) detector arrays are
commonly used in low-light imaging applications, and are known
to suffer from certain defects which cause ‘hot pixels’. These are
detectors which exhibit a significantly larger than average dark count
rate, adding noise to the data. Typically, data from these detectors
are removed by post-processing or the detectors can, in some cases,
be switched off prior to measurement. Users must define and identify
these hot pixels, however there exists no consistent methodology of
doing so. The authors present a self-consistent method of defining
a hot pixel by fitting a log-normal distribution to a histogrammed
dark-count map of the array. The approach has proven a robust
method of classifying hot pixels in a number of different detector
arrays, providing a threshold based on statistical analysis rather than
human intuition. This definition provides a reliable and standardised
figure of merit, facilitating a more accurate comparison between
different single-photon detector arrays.