To address the problems such as low efficiency and repeated calculation in multi-threshold selection for the traditional multi-threshold infrared image segmentation, a fast multi-threshold infrared image segmentation algorithm based on an improved artificial bee colony algorithm is proposed. First, the artificial bee colony algorithm is introduced in threshold selection to implement multi-threshold segmentation. Then, to overcome the shortcomings existed in the original artificial bee colony algorithm, such as low convergence speed and being easy to fall into local optimum, an improvement is made in the search by leaders, followers and scouts. The improved algorithm is faster and can converge to the optimal solution more accurately. The experimental result shows that compared with the original artificial bee colony algorithm, this improved algorithm is faster for the same accuracy and its result is closer to the optimal solution for the same iterative times. It can implement multi-threshold segmentation of infrared images very efficiently while keeping its accuracy. It is a feasible segmentation method of infrared images.