The choice of step length plays an important role in convergence speed and precision of Cuckoo search algorithm. In the paper, a self-adaptive step Cuckoo search algorithm based on dimensional improvement is provided. First, since the step in the original self-adaptive step Cuckoo search algorithm is not updated when the current position of the nest is in the optimal position, simple modification of the step is made for the update. Second, evaluation strategy based on dimension by dimension update is introduced to the modified self-adaptive step Cuckoo search algorithm. The experimental results show that the algorithm can balance the contradiction between the global convergence ability and the precision of optimization. Moreover, the proposed algorithm has better convergence speed.