一种基于方向可控金字塔结构的快速医学图像配准算法

A Fast Medical Image Registration Algorithms Based on Steerable Pyramid

提出了一种快速医学图像配准算法.采用方向可控滤波器,稳定地提取图像多方向特征信息,合成特征图来进行更精确的配准.为了避免全局搜索旋转中心造成的高运算量,提出选取特征图的质心作为旋转中心进行图像空间变换而后逐层配准的方法.另外,计算质心还可以提供平移参数的初始估计范围,加速算法搜索时间.仿真结果证明该算法的精确度高,运算时间大为减少,在多模态图像配准中也有很好的效果.

A fast algorithm for medical image registration is presented. With steerable filters, the multi-directional feature information of image can be steadily extracted and then the synthesized feature graph for more accurate registration can be obtained. To avoid heavy computation due to the globally searching rotation center, a method to determine the centroid of feature graph as rotation center for geometry transform is presented, which uses steerable pyramid for level by level registration. Furthermore, centroid calculation can also provide initial estimation of translation parameters, which greatly accelerates the search process. Simulation results prove that the accuracy of the algorithm is high and the computation time of the algorithm can be remarkably reduced. For multimodal image registration, the algorithm also leads to a good result