Abstract: Because the traditional Otsu multi-threshold image segmentation algorithm usually takes too much time to find the optimal segmentation threshold. Therefore, this paper proposes an improved sparrow search algorithm (SSA) to shorten the time cost. Based on the traditional SSA, chaos initialization strategy, adaptive weighting, reverse learning strategy, and Levy flight mechanism were introduced to perform multi-threshold image segmentation. It was compared with the image segmentation results of algorithms such as PSO, GWO, SSA and ISSA. Experimental results show that the algorithm greatly shortens the running time of the traditional multi-threshold Otsu image segmentation algorithm, and improves the accuracy of image segmentation, which has certain practical value.