介绍了一种基于最大互信息原理的图像配准技术。并针对基于最大互信息图像配准的不足,研究了基于Harris角点算子的多模态医学图像配准。在计算互信息的时候,采用部分体积插值法计算联合灰度直方图。在优化互信息函数的时候采用了改进的遗传算法将配准参数收敛到最优值附近。实验结果表明本方法具有较高的配准精度和稳定性。 An image registration technique based on the principle of maximum mutual information is introduced. In order to overcome the shortcomings of image registration based on maximal mutual information, multimodality medical image registration based on Harris corner operator is studied. When calculating the mutual information, we use the partial volume interpolation to calculate the joint gray histogram. In the optimization of mutual information function, an improved genetic algorithm is adopted to converge the registration parameters to near the optimal value. Experimental results show that this method has high registration accuracy and stability.