Comparative study of deconvolution techniques for infrared speckle interferometry

Three deconvolution techniques, CLEAN, Maximum Entropy Method (MEM), and L. B. Lucy’s algorithm, are compared through both one-dimensional computer simulation and real astronomical data obtained at high angular resolution. These algorithms are evaluated according to speed, noise sensitivity, and resolution gain. For the simulations, we define the resolution gain needed to separate the components of an equal double as the ratio of the FWHM of the point spread function to the double separation. CLEAN is at least two orders of magnitude faster than the other two algorithms but is limited in usefulness to resolution gains near unity. Lucy’s algorithm converges rapidly at first and achieves a resolution gain near two but requires smoothing techniques to reduce its sensitivity to noise. MEM is less sensitive to noise and is capable of resolution gains as high as four. These conclusions are also demonstrated with real astronomical data where MEM again produces the smoothest image consistent with the signal-to-noise.