The roughness of rock joint surfaces is the most important factor in determining their shear strength. Currently, the roughness of rock joint surfaces is mostly expressed using the root mean square slope (Z2) value, which is closely related to the sampling interval in the process of digitizing joint surfaces. To further improve the calculation accuracy of rock joint surface roughness, this paper conducted research on the characterization and strength prediction methods of rock joint surface roughness considering sampling interval. The main research results are as follows: First, the roughness of joint surfaces varies among different types of rocks after splitting, with granite exhibiting the highest roughness, followed by quartzite, and then red sandstone and mudstone. Second, the Z2 values of different joint surface contour lines show a linear decrease with the increase of sampling interval. And the greater the fluctuation of the joint surface, the more obvious the influence of sampling interval on Z2 value. A calculation formula for the joint surface roughness coefficient, which considers the sampling interval, was established based on a logarithmic function. This new formula offers clearer physical meaning and higher prediction accuracy compared to the original JRC prediction model.