Abstract: The shortage of current algorithm of White Matter Lesion Segmentation is not really can be used in realtime tasks.The main reason is the low performance characteristics to be used.To solve this problem, this paper proposes a new White Matter Lesion Segmentation feature and method of data processing to improve the performance of the White Matter Lesion Segmentation.This method uses Divide and Conquer and treats WM and GM differently.With the ACCORD-MIND data set of MRI and SVM classifier, we compare this new algorithm with traditional segmentation performance.The experimental results show that this feature is superior to the traditional features, and can obtain good segmentation results and fast segmentation speed.