关键词: 模糊[C]均值, 图像分割, 微分进化

Abstract: Fuzzy C Mean（FCM） clustering algorithm is sensitive to the initial values of the background noise and the clustering algorithm. In order to overcome this problem, a differential evolution fuzzy C means segmentation algorithm is proposed. To avoid falling into local extremum, firstly, fuzzy C means clustering is adopted to initialize, and the improved FCM is applied to carry out fuzzy clustering. In initialization population operation, the parameters of differential evolution algorithm are set, and the fitness value of each individual is calculated in the population, finally, conditions fitness value is studied to meet variation conditions, crossover and selection operation. The global search optimization ability of differential evolution algorithm can effectively restrain the effect of the background noise and texture details on the image segmentation. It also overcomes the problem that the initial value selection is sensitive to image segmentation, and it is a more efficient method to improve segmentation results. Conclusion differential evolution algorithm has the advantages of simpleness, fastness and robustness, which can effectively overcome the disadvantages of the fuzzy C means algorithm.

Key words: fuzzy C mean, image segmentation, differential evolution