Photon noise bias in computed bispectra

Summary form only given. Recent work in developing techniques for the removal of photon noise (shot noise) from two-dimensional power spectra and bispectra is reported. The general problem of speckle imaging (i.e. imaging through atmospheric turbulence), has been addressed. The general approach is to take a set of very-short-exposure frames (1-10 ms), compute some quantity like the power spectrum or bispectrum for each frame, and average this quantity over all the frames. These averaged quantities can be used to reconstruct the image. However, when the object being imaged is very dim, or the exposure time for a single frame is very short, the computed bispectrum with exhibit a photon noise component due to the Poisson statistical nature of the photon detection process. The photon noise contribution has been derived, taking into account that some pixels are brighter than others when the camera is exposed to uniform illumination.<>