Based on the SNESIM algorithm, a fast three-dimensional(3D) reconstruction algorithm of multiple-point statistics simulation was proposed. First, pattern sets were built by the red-black tree. The height of tree structures was rapidly reduced. CPU time for retrieving matching data events was saved. Then, the algorithm combined the advantages of probability aggregation approach and the sequential two-dimensional(2D) simulations with sample data approach. Half-template and all-template were utilized in nodal simulation. The proportion of informed conditioning nodes in template was increased so that the amount of possible data event was declined. The algorithm was tested on 2D channelized reservoir section, 3D porous medium images and 3D reservoir rock images. The results showed that this algorithm achieved the efficiency without degradation of accuracy. No matter isotropic or anisotropic rocks, the visual characteristics, the statistics characteristics and the pores-throats structures of the reconstructions were similar to the corresponding real 3D digital rock images, which proved that this algorithm was reliability.
