AbstractThe solution of a Fredholm integral equation of the first kind which is, in general, an ill-posed problem, can be regularized by the maximum entropy method (MEM ). With this method the solution is reformulated as a nonlinear optimization problem with one or two nonlinear constraints. In real life applications, however, this optimization problem is a large-scale one with up to 106 unknowns to be determined. For the solution of such problems we present a numerical algorithm which is made to work most efficiently on modern multiprocessor, vector computers. The usefulness of the algorithm is illustrated by deconvolving optical pictures of the sky taken with astronomical telescopes.