【摘要】 对CT和MRI图像进行配准,利用主轴法配准速度快、鲁棒性较高、易实现的特点,计算图像的初始平移量和旋转量,对图像进行粗配准。以粗配准的结果作为新的浮动图像,在基于互信息方法的图像配准中,由于目标函数产生局部极值的原因,采用模拟退火-单纯形的混合优化算法,以互信息作为相似性测度迭代搜索,使互信息最大,从而实现最佳配准。结果表明,采用由粗到细的配准策略,配准精度高,并且混合优化算法克服了单一优化算法的速度慢、易陷入局部极值的缺点,并且不需要人为调整待配准图像的分辨率,自动化程度高,配准速度快,能够满足脑图谱开发过程中的多模图像配准要求。更多还原

【Abstract】 Registration of CT and MRI images is achieved using principal axis method. This method had the characteristics of fast speed, high robustness and easy manipulation. The initial translation parameter and rotation parameter were calculated for rough registration. The results of rough registration were used as new floating image. The cause of local maximum of object functions image registration was analyzed based on mutual information and an optimization strategy was proposed by using simulated annealing-simplex method. The results show the precision is high. The new improved method or simplex-simulated annealing algorithm can prevent the optimizing process from being trapped into local extremum and low convergence speed. It has high degree of automation and the advantages of high registration speed and high registration accuracy. It can meet the demands of multi-modality image registration during the research of human brain atlas.更多还原