The threshold method is a simple and effective image segmentation technique.However,the threshold method also has obvious disadvantage,the amount of calculation for solving threshold appears to be exponential amplification with the increase of threshold.In order to overcome the shortcomings of large computation load and long computation time for multi-threshold image segmentation,we introduce an improved differential evolution algorithm,which proposes a new mutation strategy,adopts self-adaption scaling factor and cross factor,and newly adds Perturbation strategy.In order to achieve multi-threshold segmentation,the improved algorithm considers multi-threshold segmentation as an optimization problem whose objective function is formulated according to Otsu.Experimental results show that compared with other algorithms,the improved algorithm not only can achieve an accurate image segmentation result,but also has a faster speed.