一种新的基于投影矩的三维图像配准方法

New way of registration using the projection moment in 3-D image data

提出一种投影矩的概念 ,并将其用于三维医学图像的配准 .与传统的几何矩相比 ,投影矩不仅可以完备地描述图像的特征 ,而且计算量较小 ,因而可以将其用于三维图像的实时配准 .选取待配准的 2幅三维图像若干阶投影矩差的平方和为目标函数 ,采用Powell方法求取最优解 ,得到配准结果 .将方法应用于一些模拟和实际数据 ,获得了较高精度的实验结果 ;并且缩短了整个计算过程的时间 .

A new concept named projection moments, and its application to 3 D medical image registration is introduced. Compared with the method based on geometric moments, the new method needs much less computation time, so it can be used in the clinical 3 D image matching problems. A least square objective function was used to measure the difference of projection moments between the test and reference images, and the Powell method was applied to solve the optimization problem. The application of the method to some simulated and real data proved that accurate results could be obtained within a reasonable time, therefore, the new method has the potential of practical use in the medical image registration.