In order to obtain clear restoration result from a blurred image, we propose the frog leaping algorithm. The frog population is divided into several groups, each ethnic group includes a number of frogs, and every evolution only updates the worst frog′s position. In order to prevent the solution space from contraction, the best individual is chosen to perform operation with the Gaussian mutation operator. A threshold value strategy is set to guide the update of the variables in various dimensions of frog leaping. The blurred image restoration model is established, in which the relation between blurred and restored images is set up with the nonlinear mapping characteristics of the frog leaping algorithm. The blurred image restoration solution converges on the Poisson statistical maximum likelihood solution. A recovery evaluation system is established. Simulation results show that the frog leaping outperforms other algorithms with the clearer recovery, the highest improvement percentage and the least time consumption.