A two-dimensional Otsu image segmentation algorithm based on the improved firefly algorithm is proposed to solve the problem that one-dimensional maximum interclass variance algorithm has low anti-noise performance and two-dimensional maximum between-cluster variance has high computational complexity and poor real-time difference. Firstly

a fixed step factor is replaced by introducing the step adjustment function so that the step size can be obtained adaptively with the number of iterations and the convergence demand in the FA. Then

the optimization process of the 2-Otsu distance measure function is improved by a way of the improved FA which only searches for the maximum brightness value of the spot where the fireflies are located

and the optimal threshold value is obtained

and the image segmentation is performed. Experiments results show that the proposed algorithm in this paper is more stronger in the noise resistance than one-dimensional Otsu and more efficient in the partition than the two-dimensional Otsu

and on the premise that the entropy of shannon and the regional contrast keeps unchanged

the real-time segmentation time is declining about26.0% compared to the two-dimensional Otsu by the classical Lena diagram and boat diagram.