In view of the problems of dge blur, uneven gray level and high image noise presenting in traditional medical image segmentation algorithm, a medical image segmentation method was put forward based on combining the information entropy and the improved particle swarm algorithm, thus the best threshold for the medical image segmentation can be realized under the condition of ensuring the maximum information entropy. Using the maximum information entropy as the fitness function, optimal threshold segmentation of medical image was achieved by improving the particle swarm algorithm to optimize the optimal segmentation threshold. Choosing the brain images with noise or not as the research objects, intuitive and objective analysis and segmentation speed analysis indicate that, the proposed new method overcomes the defects existing in traditional segmentation algorithm to a large extent, thus the segmentation speed and accuracy are improved significantly. And also, the new algorithm has strong robustness and the ability of resisting noise.