Abstract: A method of medical image elastic registration based on contours is presented in this paper. In this method, the contour of the object is first abstracted, and then the centroid of the object is calculated. Around the centroid, the points on the contour are calculated at the same interval of angle, and these points work as the feature points. Their corresponding relation can be determined by calculating similarity. The ψ-function proposed by H.Wendland is adopted as radial based functions in the elastic transform. The transformation coefficient can be computed with the two sets of feature points. Finally, the deformed image can be registered. The experiment shows that this method is fast and effective to deformed medical image registration.

Key words: medical image, elastic registration, contour, compactly supported radial functions

摘要： 本文提出了一种基于轮廓的医学图像弹性配准方法。该方法先对医学图像进行轮廓提取，计算轮廓所围部分(目标)的质心。围绕质心每隔一定角度抽得轮廓点作为特征点。通过计算相似度确定两组特征点的对应关系。采用Wendland提出的ψ函数作为弹性变换的基函数。通过两组对应点确定变换系数。最后根据配准参数对形变图像进行变换，从而得到配准图像。实验表明，本方法对具有形变的医学图像的配准是快速的、有效的。

关键词: 医学图像, 弹性配准, 轮廓, 紧支撑径向函数