With the development of medical imaging technology, the analysis of medicalimages has become a vital component of a large number of applications. Suchapplications occur throughout the clinical track of events. Since information gainedfrom two medical images acquired in the clinical track of events is usually of acomplementary nature, proper integration of useful data obtained from the separateimages is often desired. To integrate the data, we need medical image registration.In this paper, we present a noval medical image registration algorithm based onscale-invariant salient region features. In contrast with using pure feature-based orintensity-based method, we adopt a hybrid algorithm which integrates the merits ofboth approaches.Our algorithm firstly automatically extracts scale-invariant salient regionfeatures whose interior intensities could be matched using robust similarity measures.After primarily matching the feature pairs extracted from the first step, we findoptimum transformation parameters between the two images based on geometricconstraints. This algorithm could be applied to mono-modal and multi-modalmedical image registrations and it is proven to be a robust and precise method in ourtest. |