基于改进的互信息配准方法在多模态医学图像中的应用研究  

Research on the Application of Improved Mutual Information Registration Method in Multimodal Medical Images

作　　者：何迪 HE Di(School of Information and Electronic Technology,Jiamusi University,Jiamusi Heilongjiang 154000,China)

机构地区：[1]佳木斯大学信息电子技术学院,黑龙江佳木斯154000

出　　处：《信息与电脑》2025年第2期4-6,共3页Information & Computer

摘　　要：图像配准作为计算机视觉和医学图像处理中的重要技术,被广泛应用于多模态医学图像分析中。传统的互信息配准方法通过基于图像强度的相似性度量来实现配准,但在处理具有较大灰度差异、复杂几何变形或噪声干扰的医学图像时,常常遇到局部极值和配准精度不足的问题。为此,研究提出了一种改进的互信息配准方法,它能够有效避免传统方法中的局部最优解,显著提高配准精度和鲁棒性。实验结果表明,改进后的方法在处理多模态医学图像时,特别是在图像间存在显著差异的情况下,展现了较传统方法更高的配准精度和更强的稳定性。As an important technique in computer vision and medical image processing,image alignment is widely used in multimodal medical image analysis.The traditional mutual information alignment method realizes the alignment by similarity metric based on image intensity.Still,it often encounters problems with local extremes and insufficient alignment accuracy when dealing with medical images with large grayscale differences,complex geometric distortions,or noise interference.To solve these problems,the study proposes an improved mutual information alignment method,which can effectively avoid the local optimal solution in the traditional method and significantly improve the alignment accuracy and robustness.Experimental results show that the improved method demonstrates higher alignment accuracy and stronger stability than the conventional method when processing multimodal medical images,especially when there are significant differences between images.

关 键 词：图像配准 互信息 线性插值 B样条插值 最近邻插值 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]