基于灰度互信息和梯度相似性的医学图像配准及其加速处理

Medical image registration based on grey mutual information and gradient similarity with an accelerated processing method

研究基于归一化互信息的医学图像刚性配准算法,提出改进配准速度和改善配准精度的相应措施.配准处理包含3项主要计算处理,即空间变换、互信息计算以及优化搜索.针对不同计算处理分别研究了相应加速策略,提高其计算速度,实现三维体数据的快速配准.并且,针对传统基于互信息测度配准方法未利用图像灰度空间分布信息,提出将灰度变化梯度相似性与互信息相结合的配准方法,从而进一步提高了配准算法的精度和鲁棒性.实验结果表明了算法的有效性.

This paper presents new methods that have been developed for rigid registration of medical images. These methods are based on normalized mutual information and improve registration speed and precision. The whole regis- tration process includes three main steps： space transformation, mutual information calculation, and optimal search. Some acceleration strategies for fast registration of 3D volume data were investigated. Conventional registration approaches, based on mutual information, neglect the spatial distribution information of images. In view of this drawback, a new method was developed, combining data based on gradient similarity and mutual information. This improves the precision and robustness of the registration algorithm. Experimental results proved the validity of the proposed methods.