Medical image registration: Comparison and evaluation of nonlinear transformation algorithms

Abstract

Image registration seeks to compare and combine images acquired from multiple modalities, at different time or at different viewpoints by feature based approach or optimizing the similarity measure of two image sets. In the landmark based registration, the transformation function is required to spatially match the features. Image guidance systems designed for neurosurgery, hip surgery, and spine surgery often relies on feature based registration. Accuracy is important to these systems. In this paper, the transformation functions like polynomial, piecewise linear (PL), local weighted mean (LWM) and thin plate spline (TPS) are evaluated. The comparison will be made in terms of the registration time, error rate, correlation index and degree of matching.