Abstract: Aimed at the shortcomings of high computation cost and poor accuracy of the multi-level thresholding image segmentation, an improved whale optimizing Otsu multi-level threshold image segmentation algorithm integrated with self-learning and multi-leader is proposed. In order to improve the optimization precision and the convergence rate of traditional WOA, the multi-leader strategy with memory mechanism was introduced to enhance the global search ability of the population and avoid a local optimum in late iterations. The individual self-learning mechanism for leads was designed to promote the population diversity. The Levy flight mechanism was used to improve the robustness of the algorithm and avoid the premature convergence, so that we implemented an improved whale optimization algorithm MLWOA. Using Otsu between-class variance as fitness function, MLWOA was used to search the optimal multi-level threshold group of image segmentation for determining the optimal segmentation thresholds. By comparing some index such as the peak signal-to-noise ratio PSNR, the structural similarity SSIM and the feature similarity FSIM, the obtained experimental results verify that our algorithm can obtain a higher segmentation accuracy and a higher segmentation efficiency than the same kinds.