中文摘要：

弹性配准是医学图像处理的重要研究内容，为了精确得到图像的局部几何变形，需要大量均匀分布的标记点来表示局部偏差，为此，提出了一种自动、准确的基于区域相似性的标记点选取方法，并利用层次B样条散乱数据插值方法能够解决逼近精度和光滑性的平衡问题的特点，建立一种新的配准方法．实验证明，本文提出的自动标记点的选取方法能够产生准确的表示局部变形的标记点对应性，所建立的配准方法是一种准确、简便和稳定的方法，能够产生光滑、准确的配准变换．

英文摘要：

Elastic medical image registration is a most the local geometric deformation of image precisely, it important issue in medical image processing. To obtain requires large numbers of evenly distributed landmark points to represent local displacements. An automatic and accurate method to extract landmark points is proposed based on regional similarity between the two images. Furthermore, by utilizing the characteristics of multilevel B-splines to balance approximation precision and smoothness, a new registration method is presented. Experimental results show that the proposed landmark points extraction method can produce accurate point correspondence presenting local deformation, and the new registration method is accurate, convenient and stable which can obtain smooth and accurate registration transformation.