Multimodal Brain Image Registration Based on Wavelet Transform Using SAD and MI
Multimodal Brain Image Registration Based on Wavelet Transform Using SAD and MI
复制标题
基于小波变换的SAD和MI多模态脑图像配准
DOI：
10.1007/978-3-540-28626-4_33
复制
发表时间：
2004
影响因子：
0.7
通讯作者：
Albert C. S. Chung
中科院分区：
文献类型：
--
作者：
Jue Wu;Albert C. S. Chung
多分辨率方法通常用于加速基于互信息（MI）的配准过程。传统上，高斯金字塔通常用作多分辨率表示。然而，在多模态医学图像配准中，基于高斯金字塔的MI方法可能遭受捕获范围短的问题，特别是在较低的分辨率水平下。提出了一种基于小波变换的多模态图像配准方法，该方法采用了两种匹配准则，即差和准则（sum of difference，SAD）用于提高配准的鲁棒性，MI准则用于保证配准精度。实验结果表明，该方法在保持相当精度的同时，获得了比传统的基于MI的高斯金字塔方法更长的捕获范围。
The multiresolution approach is commonly used to speed up the mutual-information (MI) based registration process. Conventionally, a Gaussian pyramid is often used as a multiresolution representation. However, in multi-modal medical image registration, MI-based methods with Gaussian pyramid may suffer from the problem of short capture ranges especially at the lower resolution levels. This paper proposes a novel and straightforward multimodal image registration method based on wavelet representation, in which two matching criteria are used including sum of difference (SAD) for improving the registration robustness and MI for assuring the registration accuracy. Experimental results show that the proposed method obtains a longer capture range than the traditional MI-based Gaussian pyramid method meanwhile maintaining comparable accuracy.