Discontinuity sets play an essential and pivotal role in the deformation monitoring and stability analysis of the rock mass, but there are still many challenges for accurately and rapidly extracting discontinuity. In this study, an extraction and characterization method of discontinuity sets based on point cloud supervoxel segmentation was proposed, which consists of four parts: 1) a multiresolution supervoxel segmentation (MRSS) algorithm was developed to classify unstructured point cloud into multiresolution facets and discrete points; 2) to extract the individual discontinuity, the single supervoxel that having spatial connectivity, similar planarity, and parallelism was clustered; 3) the orientation of individual discontinuity was calculated, respectively, based on the plane fitting parameters; and 4) for comprehensively analyzing the stability of rock mass, the improved K -means clustering algorithm is utilized to constructing the discontinuity sets that having similar orientation information. The novel method has been successfully tested on two practical cases (a rock cut and a side slope point cloud captured by the terrestrial laser scanner). A comparison with existing methods shows that the deviation of the discontinuity orientation for rock cut is less than 1°, and the time efficiency is increased by 2.6 times. In addition, the orientation variation of the seven principle discontinuity in the five temporal side slope point cloud is relatively small, the dip direction and angle are within 2° and 1°, respectively. We can conclude that the proposed method can efficiently obtain the full extent of every individual discontinuity from rock mass surface point cloud and accurately analyze their orientation information.