An improved stored grain insect image segmentation of tower layered - fuzzy C-means algorithm（TL-FCM）was brought out to solve the problems such as high operation cost and high noise sensitivity in traditional fuzzy C-means algorithm (FCM). A tower layered structure was adopted to reduce computation time complexity, at the same time to restrict the target image subordinate degree of pixels. The constraint was added into the objective function of traditional algorithm, therefore neighborhood information was effectively constrained. Simulation results showed that the new algorithm needed less computation time, effectively retained detail image information with ideal performance and results.