ABSTRACT
As reviewed in (Zitová and Flusser 2003), fourmain deformable registration techniques were developed for medical image data: elastic registration, level-set method, diffusion-based registration, and optical flow method. In elastic registration (Bajcsy and Kovacic 1989; Davatzikos, Prince, and Bryan 1996), external forces are introduced to stretch the image while internal forces defined by stiffness or smoothness constraints are applied to minimize the amount of bending and stretching. One of its advantages is that the
feature matching and mapping function design can be done simultaneously. The level-set method (Osher and Fedkiw 2003) is a numerical technique for tracking interfaces and shapes, which can easily track topology change and combine segmentation together with registration (Moelich and Chan 2003; Droske and Ring 2006).The diffusion-based registration (Thirion 1998; Andresen and Nielsen 2001) considers the contours and other features in one image as membranes, and the other image as a deformable grid model, with geometrical constraints. This approach relies mainly on the notion of polarity, as well as the notion of distance. The optical flow method (Horn and Schunck 1981; Barron, Fleet, and Beauchemin 1994) assumes that the corresponding intensity value in the static image and the moving image stays the same, and then estimates themotion as an imagevelocity or displacement. This method is suitable for deformations in temporal sequences of images. Optical flow and diffusion registrations can be combined to have better matching results.
