基于QPSO的多模态医学图像配准研究  

Research on Multimodality Medical Image Registration Based on QPSO

作　　者：占跃华[1] 张方[2] 

机构地区：[1]九江职业技术学院,江西九江332007 [2]华东交通大学信息工程学院,江西南昌330013

出　　处：《微计算机信息》2010年第10期150-152,共3页Control & Automation

摘　　要：本文提出一种基于QPSO的形态金字塔多模态医学图像配准算法,该方法以梯度互信息作为相似性度量标准,运用全局仿射变换建立数学模型,充分利用形态金字塔的非线性、多分辨率等特性将图像进行分层处理,并且利用QPSO和Powell算法结合逐层获取参数解。实验结果表明在任意初始值的情况下,基于QPSO的寻优算法比Levenberg-Marquardt寻优算法运行速度快,并且配准精度更高。We proposed a morphological pyramid multi-modal medical image registration algorithm based on Quantum Particle Swarm Optimization （QPSO）.This approach utilizes the global affine transformation model between images with the non-linear and multiresolution properties of the morphological pyramid structure,and bilinear interpolation.This proposed algorithm can be implemented hierarchically and obtain the optimal parameters on each layer of the pyramid with QPSO.In the process of registration,the gradient of mutual information is used as similarity metrics which can effectively improve the alignment accuracy.Digital simulations show that this optimization algorithm works with higher speed and precision than Levenberg-Mardquradt（LM） optimizing algorithm.

关 键 词：形态金字塔 多模态 图像配准 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]