Non-rigid image registration based on Overlapped Block Check and Free-Form Deformation

A new non-rigid registration approach is presented for the images which contain highly local deformation, using Free-Form Deformation (FFD) and Overlapped Block Check method. First of all, the hierarchical B-spline FFD with large grid is used for global registration. Then the rough registered image and fixed image are divided into a series of corresponding overlapped block pairs. We calculate the difference of each block pair and decide which ones need further refined registration, then a small grid FFD is adopted for them. In order to avoid the boundary effect, we propose a strategy to regulate the motion of the control points near the boundaries. Experimental results on Lena image and Brain image show that the proposed method can achieve high registration accuracy not only through difference images between registered images and fixed images, but also by similarity measures, such as Mutual Information (MI), Normalized Mutual Information (NMI) and Sum of Square Difference (SSD), etc. Furthermore, we confirm that the large and small spacing of control points in FFD will influence the accuracy of image registration in different ways.