Abstract: In the progress of 3-Dimensional (3D) reconstruction from 2-Dimensional (2D) rock slice images, it is difficult to determine the auto-correlation function of the 3D microstructure. The reconstruction method based on the fast Fourier transforms uses empirical formula to predict the auto-correlation function of the 3D microstructure, but the reconstruction result has relative large error. Another reconstruction method based on Particle Swarm Optimization (PSO) to optimize the reconstruction progress is proposed in this paper. This method sets autocorrelation function of 3D microstructure as the position of particles, calculates fitness value as the error between 3D microstructure and 2D image. Compared with FFT reconstruction method, the similarity between reconstruction results and 2D image is greatly improved. Compared with simulated annealing method, the proposed method reaches similar reconstruction result. The reconstruction method based on PSO is more efficient and can be well applied to the image reconstruction.