Abstract: The effects of multi-resolution approaches to rigid registration based on mutual information are discussed，aimming at an acceleration of the matching process while maintaining the accuracy and robustness of the method and the strong anti-noise.The behavior of mutual information matching in a multi-resolution scheme is examined for magnetic resonance and computed topography images.The normalized mutual information and the attribute scale space are combined in registration for the first time.The experiments show that the multi-scale applying in mutual information matching is an appropriate method.

Key words: register, normalized mutual information, attribute scale space, rigid transformation

摘要： 提出基于多尺度图像的互信息测度的配准方法，主要讨论这种方法的性能。提出这种方法的主要目的是在保证这种方法的准确性和鲁棒性的同时加快配准速度，并且具有强抗噪性。将这种方法在MR图像和CT图像上进行检验。这是首次将属性尺度空间与互信息结合，实验表明多尺度图像的互信息测度是一种行之有效的配准方法。

关键词: 配准, 归一化互信息, 尺度空间, 刚体变换