基于刚-弹性变换的医学图像塔式配准新算法

A new mixing rigid-elastic multiresolution algorithm for medical image registration

提出了利用B样条实现医学图像从刚性到弹性配准的统一的塔式算法。选取图像的四个顶点作为构成变形函数的B样条的节点,合理选择B样条的次数,可以实现图像的仿射变换。沿X和Y方向均匀增加节点数量,逐步增加变形函数的复杂性,通过选择B样条的不同的次数n,可以实现图像的分块仿射变换或n-1阶导数连续的弹性变换。整个匹配算法体现了从整体到细节的匹配思想。实验证明,这种配准方法比单纯的弹性配准方法鲁棒性有很大提高。

We present a new mixing rigid-elastic multiresolution algorithm for medical image registration from global registration to local registration in sequence. The global registration is achieved by the method of affine transformation composed ofB spline, which knots are the four vertexes of the medical image. When increasing the number of the knots of the B spline along the X and Y axis, the transformation function will be more complex and ＂elastic＂, which can complete the elastic aligning for the detail of the medical image. The results of experiment demonstrated that the presented algorithm was more robust than the previous elastic algorithm of registration.