Brain MRI images often have a lot of noise, and the edge is not clear, which makes the traditional fuzzy C-means (FCM) clustering algorithm can not obtain accurate brain tumor segmentation results. Therefore, a brain tumor image segmentation method based on Markov random field (MRF) and hybrid kernel function clustering is proposed. Firstly, particle swarm optimization is used to initialize the cluster center; Then, the single Gaussian kernel function in the traditional kernel fuzzy clustering algorithm (KFCM) is replaced by the mixed Gaussian kernel function; Finally, the prior probability of Markov random field is introduced to modify the objective function of the algorithm and enhance the anti noise performance of the algorithm. The experimental results show that the proposed algorithm has good anti noise performance in brain tumor image segmentation, and the segmentation accuracy is significantly higher than the traditional algorithm. The average values of dice index and Jaccard index are 0.9501 and 0.9051 respectively.