PDF version of this document

INTRODUCTION

W e embark on the task of assessing non-rigid registration (NRR) and we have a group of methods for doing so. For each dataset that we are given, we are able to assign a measure, which is representative of registration quality. That measure, however, is subjected to some level of uncertainty, thus it must be treated as a value with a certain variance. This document outlines the way in which we derive information about the standard errors. This should include inter-instantiation error, which is present when repeated experiments are performed and averaged over. We also deal with some error that is associated with the calculation of registration quality, which is either model-based or overlap-based in our case.