Estimation of spatial power spectra in speckle interferometry

Abstract
Calculations of the signal-to-noise ratio (SNR) of estimates of the power spectra of spatially varying random processes, such as stellar speckle patterns, usually include realizations that contain less than two detected photons. It is shown in this paper that if these cases are excluded from the analysis procedure either implicitly or explicitly, then under the usual definition of SNR, the overall SNR of an estimate can increase by up to a factor of N¯-½., where N¯ ≪ 1 is the average number of detected photons per realization.