STATISTICAL IMAGE RECOVERTY

Abdulkarim M. HAMMOUD

It is desired to estimate the intensity of a self-luminous object in the presence of atmospheric turbulence. The object emits quasimonochromatic, incoherent light that is observed by an imaging system with a known intensity point-spread function. Data are collected using a photon sensing or a photon counting (CCD) camera. In addition to photon noise, collected data suffer from background radiation, and camera non-ideal characteristics in the form of dark current and non-uniform flat field response. Depending on the statistical description of the motion process, a set of iterative, maximum likelihood based algorithms are derived to estimate the object intensity. The derived algorithms compensate for the effects of random motion, imaging system point-spread function, non-zero background radiation, and camera non-ideal characteristics.

An algorithm for image recovery from n^th order correlation functions is derived. Image recovery from second and third order correlation functions are presented as a special case of the above algorithm. These algorithms are used in conjunction with speckle interferometric techniques to correct for general atmospheric turbulence us­ing simulated and real data. In general, random translations and 180 degree rotation ambiguity is not present in the estimate obtained using the above algorithms.

In addition, image recovery algorithms that are widely used in tomography, optical sectioning microscopy, and astronomical imaging are shown to be a special case of a derived general algorithm. The derived algorithms implicitly enforce the image non-negativity and support constraints.