Abstract: This paper studied the medical image registration method which was based on mutual information similarity. In the mutual information calculation process, the data of images were sampled by an entropy-based sampling method. This method divided the image into small blocks and calculated the entropy of each. The blocks were classified according to the size of entropy and different category has different sampling factor. The blocks with higher entropy should be sampled with higher factors while the blocks with smaller entropy should be sampled with smaller factors. The experimental results prove that this method can not only ensure the accuracy of registration but also accelerate the speed of registration, and it is suitable for medical image registration of real-time processing.

Key words: mutual information, similarity measure, image registration, sample factor, information entropy

摘要： 研究了以互信息为相似性测度的医学图像配准方法，在互信息计算过程中，对图像数据的采样提出了一种基于信息熵的采样方法。这种方法是将图像分成一定数量的小方块，计算每一小方块的熵，根据熵值的大小对方块进行分类，不同的类设置不同的采样因子:熵值大的方块对应的采样因子大，熵值小的方块对应的采样因子小。通过实验证明，该方法能够折中配准的精度和速度，适用于医学图像配准的实时处理。

关键词: 互信息, 相似性测度, 图相配准, 采样因子, 信息熵