Abstract: It is difficult for completely automatic medical image segmentation of the blurred images caused by the continuity of organism.An interactive segmentation algorithm based on Radial Basis Function（RBF） curve fitting is proposed.Firstly，it allows user to interactively select some knots on the interested areas as interpolation centers.Secondly，an implicit surface is reconstructed from the selected knots with RBF.The iso-line of the implicit surface is the segmented areas.Two acceleration methods are suggested from the bounding box of the feature points and mathematical morphology operators.Lots of experiments show that this approach are efficient and can deal with very blurred medial images.

Key words: medical image segmentation, radial basis function, curve fitting

摘要： 针对由于机体组织间的连续造成医学图像区域边界的模糊性，导致完全自动的医学图像分割算法往往不能取得满意分割效果的问题，提出了一种基于径向基函数曲线拟合的交互式模糊医学图像分割算法。首先由用户在感兴趣的区域交互地选取一些特征点，然后利用径向基函数构造一个三维空间的隐式曲面，使该曲面上的某一等值线即为分割出的区域，文中还提出了两种不同的加速算法。大量实验表明，该算法能有效处理非常模糊的医学图像。

关键词: 医学图像分割, 径向基函数, 曲线拟合