ABSTRACT
In spline based registration (Jia et al. 2015), (Leng et al. 2013) and (Szeliski and Coughlan 1994), a spline-based control grid overlays the image, and a spatial transformation is constructed to deform this grid. An efficient and stable algorithm was suggested by (Jia et al. 2015), in which the spatial transformation was based on B-spline basis functions, and the control points were computed dynamically to enable non-rigid registration. Moreover the method was made more efficient by computing the spatial transformation on a set of successively refined grids. In our work, instead of uniformly refining the grid, we employ an adaptive mesh refinement scheme using
1 INTRODUCTION
Image registration is an image processing technique, which has a broad range of applications, but is most prominently applied in the field of computer vision and medical image analysis (Goshtasby 2005), (Zhang et al. 2012). It involves the alignment of two or more images so that there is proper integration of useful data from separate images. The important features from each of these images can be combined together to achieve a better understanding of the complete picture of the image data that is being studied. There have been many methods, by which image registration is carried out. A comprehensive survey of several such methods with their respective advantages and limitations has been covered in detail in (Maintz and Viergever 1998) and (Zitova and Flusser 2003).
