Abstract
The present paper studies the so called deep image prior (DIP) technique inthe context of inverse problems. DIP networks have been introduced recently forapplications in image processing, also first experimental results for applyingDIP to inverse problems have been reported. This paper aims at discussingdifferent interpretations of DIP and to obtain analytic results for specificnetwork designs and linear operators. The main contribution is to introduce theidea of viewing these approaches as the optimization of Tiknonov functionalsrather than optimizing networks. Besides theoretical results, we presentnumerical verifications for an academic example (integration operator) as wellas for the inverse problem of magnetic particle imaging (MPI). Thereconstructions obtained by deep prior networks are compared with state of theart methods.