LEAST-SQUARES IMAGE-RESTORATION

A general approach to the image restoration problem is presented on the basis of the method of least squares. Inverse filtering, minimum mean-square-error restoration, constrained least squares restoration, and the maximum a posteriori restoration with normally distributed errors of the pixel measurements are special cases of it. Positivity constraints are included for the object and the point spread function, which both are introduced by deterministic model functions The optimum choice of these functions is discussed, and a dynamical smoothing procedure is introduced. The algorithm opens a large field of applications including image restorations without deterministic knowledge about the point spread function and the object, or with spatially varying point spread functions. © 1979.