An information-theoretic approach to analyzing CLEAN

Abstract

We analyze the deconvolution technique, CLEAN, from an information-theoretic perspective. The CLEAN algorithm is an iterative technique, which subtracts out the target mass from the dirty image at each stage. However, each iterative step also alters the information content in the image. We investigate how the information content, measured in terms of the differential entropy, varies during the successive steps. Closed-form expressions have been derived and simulations have been carried out to corroborate the theory. It is shown that the image entropy is useful as an additional input for determining the stopping criterion for CLEAN.