Blind point-source image restoration using subspace techniques

Blind point-source image restoration refers to the problem of high resolution recovery of point-like sources which are blurred by an unknown point spread function (PSF). Applications include astronomical star field localization, magnetoencephalogram brain current imaging, and seismic deconvolution. This paper shows that with suitable constraints the problem can be cast in the language of subspace decomposition as used in blind signal copy algorithms for digital wireless communications. Assuming a separable PSF, we propose a deterministic, non-iterative ESPRIT-like solution to the restoration problem. We next extend the algorithm to non-separable PSFs by approximating them as a series expansion of a few separable components