基于概率和引力优化模型的医学图像配准

A Medical Image Registration Method Based on Probability and Gravity Optimization Model

基于互信息的配准方法,其目标函数经常存在许多局部极值,给配准的优化过程带来很大困难。提出一种基于概率模型的引力优化算法,在空间中随机构造参考物体与浮动物体,根据牛顿万有引力定律,搜索空间中质量最大的物体。利用该算法,实现以归一化互信息为相似性测度的医学图像配准实验。实验结果表明,这种方法能够有效地克服互信息的局部极值,在配准精度、配准时间和抗噪性方面都有较好的性能。

There are lots of local maximums in image registration based on mutual information,which obstruct optimization in registration process.In this paper,a new optimization algorithm,called probability and gravity optimization,was proposed.We constructed reference objects and floating objects in space,each object was located randomly,then searched the object whose quality was the heaviest according to Newton′ s law of universal gravitation in the whole space.The new method was applied to medical image registration based on normalized mutual information.Experimental results showed that this registration method could efficiently restrain local maxima of mutual information function and had better performance at registration accuracy,registration rate and noise immunity.