PDF version of this entire document

Motivation

First and foremost, what motivates registration is the need to compare images of objects. If the images (and hence the objects within them) can be properly aligned, then it is possible to study the differences between the imaged objects.

The importance of registration is demonstrated by its ubiquity. It has entered several domains where reliable acquisition of fully aligned images cannot be assured [,] or relationships between images turn out to be complex. The difficulty of this problem is even greater when alignment of large groups of images needs to be achieved []. There is room for ambiguity - and consequently - misinterpretation. NRR brings together large amounts of data, enabling the examination of variation.

Misalignment in images can result from movement of subjects or objects of interest, change in view-point, or changes to general conditions at the acquisition. Misalignments can also be artifacts of morphological changes or physical abnormalities that are due to change in mass and elasticity of organs [].