Performance Analysis for Image Super-Resolution using Blur as a Cue

A number of algorithms for image super-resolution using multiple images, have been developed over the last two decades. On the other hand, a very less amount of efforts have been made to explore the issues regarding performance analysis of these methods. Since the problem of super-resolution is often a parameter estimation problem, the Cramer-Rao bound proves to be useful tool in analyzing the performance of the estimators. We focus on the problem of super-resolving with blur as a cue. In this paper we look at the factors affecting the achievable bounds in super-resolution. We analyze the effects of the magnification factor, modeling noise and the spectrum of the signal to be super-resolved.