为了对自组织映射(Self Organizing Map, SOM)图像分割方法进行加速，提出将原串行方法延伸到并行方法，并针对特定图像分割方法设计相应的并行计算模块。为了减少主要计算负荷，通过最小化非边缘模式子块矢量的类内协方差矩阵迹到类间协方差矩阵迹的比率，计算比率函数；通过类内散射矩阵到类间散射矩阵的比率，以估计分割数量。人脑核磁共振影像(MRI)的分割结果表明：所提并行优化方法与源串行方法相比，在保证图像分割质量的前提下，大幅提高了计算效率，同时其计算效率也高于同类并行加速方法。

To accelerate the segmentation method of self organizing map (SOM), the original serial method is extended to the parallel method, and the corresponding parallel computing module is designed for the specific image segmentation method. In order to reduce the computational load, by minimizing the non block edge pattern vector in-class covariance matrix to the ratio between class covariance matrix, the ratio function is calculated. And the number of segmentation is estimated by the ratio between in-class scatter matrix and class scatter matrix. Segmentation of brain magnetic resonance imaging (MRI) images shows that the proposed parallel optimization method has greatly improved the computational efficiency under the premise of ensuring the quality of image segmentation in comparison with the original serial method. And its computing efficiency is higher than that of similar parallel acceleration method.