基于MRF模型的多模态图像配准技术研究  

Research on multimodal image registration technology based on MRF model

作　　者：袁桂霞[1] 

机构地区：[1]江苏开放大学信息与机电工程学院,江苏南京210017

出　　处：《现代电子技术》2018年第1期57-61,66,共6页Modern Electronics Technique

摘　　要：在医学图像处理领域中,医学图像配准技术极其重要,其价值体现在临床医学中对图像处理技术的应用。在解决多模态图像配准的相关问题时,基于互信息方法的应用最广泛,但在某些特定的应用中该方法受到的约束仍然较多。针对这一情况,提出一种新的医学图像配准算法,模态变换的引入作为此研究算法的基础,之后新的马尔可夫能量函数则根据两幅通过模态变换后的图像矩阵以及原配准图像得以构建。同时,为了优化能量函数引入了一种改进的梯度下降算法,从而得到配准结果。最后,运用不同的医学图像进行配准实验来验证该算法,通过实验证明该配准算法具有良好的有效性及抗噪性能。In the field of medical image processing, the medical image registration technology is extremely important. Its value is embodied in the application of image processing technology in clinical medicine. For solving the relevant issues of multimodal image registration, the method based on mutual information is most-widely used, but constrained in some specific applications. In view of this situation, a new medical image registration algorithm is proposed. The modal transformation is introduced as the basis of the studied algorithm. After that, the new Markov energy function is constructed according to the two image matrixes after modal transformation and original registration image. In order to optimize the energy function, an improved gradient descent algorithm is introduced to get the registration results. The algorithm was verified with registration experiments of different medical images. The experimental results show that the registration algorithm has perfect effectiveness and high anti-noise performance.

关 键 词：图像配准 模态变换 马尔可夫随机场 梯度下降算法 非刚体配准 图像矩阵 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP311.17[自动化与计算机技术—计算机科学与技术]