摘要
为了提高配准的精确度和鲁棒性,本文提出了一种新的基于结构信息的多模医学图像配准.该方法首先分别求出参考图像和浮动图像的相位一致图像和梯度幅值图像;其次将相位一致图像和梯度幅值图像相结合,得到了多模图像的结构信息;最后利用遗传算法和区域互信息对经过上述处理的图像进行配准.通过在MRI和CT图像上的对比实验表明:相对于经典的基于最大互信息的配准方法,本文配准方法能在更少的迭代次数内得到误差更小的配准参数,且对于噪音环境具有更强的适应能力.因此对于多模医学图像配准,本文方法是一种较传统最大互信息法更为有效的配准方法.
In this paper,a new multimodality medical image registration algorithm that based on structure information is developed to improve the accuracy and the robustness of registration.Firstly,the phase-matched images and the gradient magnitude images of the floating and reference image are obtained respectively.And then,the structure information was constructed by the phase congruency images and the gradient magnitude images.Finally,the genetic algorithm and regional mutual information were used to register the processed images.Experiments on MRI and CT show the proposed algorithm can gain the more accurate registration parameters within less iteration,and its adaptability under noise environment is also improved.Compared with the traditional mutual information algorithm,our algorithm is more efficient in the registration of magnitude medical images.
出处
《测试技术学报》
2016年第4期284-291,共8页
Journal of Test and Measurement Technology
基金
国家自然科学基金资助项目(61462063)
江西省自然科学基金资助项目(20151BAB205050)
江西省教育厅基金资助项目(GJJ14503)
关键词
图像配准
相位一致
梯度幅值
区域互信息
image registration
phase-matched
gradient magnitude
regional mutual information