摘要： 提出了一种脑功能磁共振图像配准的方法。Legendre 正交矩可以用来作为图像配准的策略，它的快速计算至关重要，边界的拟合精度和速度对Legendre 矩的计算影响很大。根据推导出的Legendre 矩的边界特点，提出了采用四连通链码法和改进的矢量斜率法进行边界拟合，从而解决了Legendre 矩快速计算的问题。采用遗传算法进行多参数配准策略优化，避免了局部极值的干扰。改进的Legendre 矩配准方法是一种快速脑功能图像配准方法。

关键词: Legendre 矩；曲线拟合；图像配准

Abstract: This paper introduces a registration method for fMRI. In order to use the Legendre moment as registration strategy, the speed of Legendre moment calculation and the boundary linear approximation are very important. Considering the character of the Legendre moment, it gives 4-connected chain code and improves vector slope to fit the boundary, which can solve the calculation of Legendre moment. The paper uses the genetic algorithm to optimize the multi-parameter strategy based on Legendre moment, which can’t be influenced by the local optimization. The experimental results show the accuracy of the method and feasibility in fMRI registration

Key words: Legendre moment; Linear approximation; Image registration