An iterative deautoconvolution algorithm for nonnegative functions

Abstract

This paper considers the inverse problem of recovering a nonnegative function from its autoconvolution. We propose an algorithm that solves the problem by minimizing Csiszár's I-divergence between the observed autoconvolution and an estimated autoconvolution. We call it a deautoconvolution algorithm. Various properties of the algorithm are discussed and proven. The effectiveness of the algorithm is illustrated via numerical experiments.