Shape and radiance estimation from the information divergence of blurred images

Abstract

We formulate the problem of reconstructing the shape and radiance of a scene as the minimization of the information divergence between blurred images, and propose an algorithm that is provably convergent and guarantees that the solution is admissible, in the sense of corresponding to a positive radiance and imaging kernel. The motivation for the use of information divergence comes from the work of Csiszár [5], while the fundamental elements of the proof of convergence come from work by Snyder et al. [14], extended to handle unknown imaging kernels (i.e. the shape of the scene).