自适应层次B样条弹性图像配准方法

Adaptive hierarchical B-splines elastic image registration method

为了准确、快速地得到图像的局部变形,基于自由变形思想,提出了一种自适应层次B样条弹性配准方法。使用由粗到精的方式,估计全局非线性变形,根据每一个标记点的匹配误差来进一步确定局部变形区域;对该变形区域进行细化,得到更精确的配准函数,直至所有标记点的匹配误差均在设定的阈值范围内。该方法使得每一层都比前一层更加接近实际的图像变形,保证了配准变换的精确性、光滑性。对MRI脑图像进行配准实验,实验结果表明了该方法准确、高效、稳健性强,特别适用于具有复杂局部变形的图像。

To get the accurate local transformation of images registered quickly,a new elastic registration method based on the free-form deformation was proposed.This method adopted the coarse-to-fine strategy,initially the global nonlinear transformation was estimated, and then the regions with local deformation was determined on the basis of the matching error of each landmark point and they were re-fined.Accordingly,the control lattice resolution in these regions was increased and more accurate registration function was obtained. This process was performed until the errors of the entire landmark points were below the predefined threshold.Using this method,each level approached a more practical deformation than the former level,which assured the registration transformation was accurate and smooth.The registration experimental results for the MRI brain images showed the proposed method was an accurate,fast and robust elastic image registration method,especially for the images with complex local deformation.