Next: Unidirectional Linear-elastic Registration algorithm Up: index Previous: Transitivity property

Image Registration

A unidirectional [2,20,21] and an inverse-consistent [16,22,17] linear-elastic image registration algorithm were analyzed to determine the degree of inverse consistency error and transitivity error that they produce.

In general, unidirectional image registration algorithms that are not landmark based do not produce transformations that have the inverse consistent property. Furthermore, unidirectional landmark registration algorithms only produce inverse consistent transformations at the landmark locations. The consistent linear-elastic image registration algorithm[16,17] was developed to overcome this problem. In this method, the forward and reverse transformations between two image volumes are jointly estimated while constraining the forward and reverse transformations to be inverses of one another. The results section of this paper demonstrate that the consistent linear-elastic image registration algorithm produces pairwise transformations that are much closer to the identity map than the unidirectional linear-elastic image registration algorithm.

Subsections

Gary E. Christensen 2002-07-04