Abstract: In view of the low separation performance of traditional independent component analysis methods for blind source separation, a blind source separation method based on improved lion swarm optimization is proposed and applied to image blind separation. On the basis of the original lion swarm optimization, the optimization combined with the strong local search ability of butterfly optimization algorithm and the excellent evolutionary mechanism of immune concentration selection, and adjusted the search balance of the algorithm through the inertia weight based on vector distance. The algorithm took the negative entropy and kurtosis of the signal as the objective function, and realized the blind separation of mixed signals by solving the objective function. Simulation results show that the proposed algorithm can effectively separate noisy mixed images, has better separation performance than the contrast algorithm, and has better separation performance under the kurtosis based objective function.