Feature-Based Registration Techniques
Christian D. Lorenz, Tobias Klinder, Jens von Berg · Biological and medical physics series · 2013
In contrast to intensity-based image registration, where a similarity measure is typically evaluated at each voxel location, feature-based registration works on a sparse set of image locations. Therefore, it needs an explicit step of interpolation to supply a dense deformation field. In this chapter, the application of feature-based registration to pulmonary image registration as well as hybrid methods, combining feature-based with intensity-based registration, is discussed. In contrast to pure feature based registration methods, hybrid methods are increasingly proposed in the pulmonary context and have the potential to out-perform purely intensity based registration methods. Available approaches will be classified along the categories feature type , correspondence definition , and interpolation type to finally achieve a dense deformation field.