Abstract

The triple correlation (TC) algorithm, or bispectrum method, has proved to be a robust and reliable technique for recovering atmospherically perturbed, stellar spatial spectra up to the diffraction limit of large telescopes. At very low light levels where the number of detected photons per image is of the order of 10², images are recorded as lists of photon addresses, and computational efficiency can be achieved by exploiting the binary nature of the photon events. In this paper we report experience with the photon address, subplane (PAS) implementation of the TC algorithm. Because the variance of the bispectrum phases cannot be readily measured in the PAS approach, the weighting factors needed when averaging multiple estimates of object phase must be approximated, and this can lead to inaccuracies. We describe a relaxation algorithm which uses the bispectrum phases as the measured constraints. This procedure gives significantly improved object phase estimates compared to the usual recursive algorithm applied to the bispectrum phases derived from the PAS process, especially in the case of real data.

© 1989 Optical Society of America