<p>The existing quantitative evaluation methods of joint roughness are rich in achievements, among them most methods mainly focus on the influence of either the inclination or the height of joint asperity, and seldom consider both the two factors in the overall roughness quantification of rock joints. In this study, a batch of granite and sandstone joints were prepared for three-dimensional morphology analysis and were subjected to direct shear tests. The Structure-from-Motion (SfM) photogrammetry technique was adopted to digitally reconstruct the three-dimensional joint morphology, which provided data basis for roughness quantification. The non-stationary joint morphology was identified and was removed from the original morphology. Based on the stationary morphology feature, the influence of asperity amplitude on roughness estimation was investigated and a significant effect was revealed. A new statistical roughness parameter, the amplitude-weighted average asperity inclination <i>θ</i>_aw, was proposed that consider extra the contribution of the asperity height feature. In order to quantitatively estimate the JRC for a certain rock joint, a prediction model was suggested to assess the JRC of two-dimensional (2D) joint profile using the new roughness parameter <i>θ</i>_aw. The model was validated to be effective in predicting JRC of joint profiles from published studies. Subsequently, another model was established to predict the JRC of three-dimensional (3D) joint surface by incorporating a three-dimensional influence factor <i>f</i>_3D into the 2D model. This 3D model was verified to have high prediction accuracy in quantifying the roughness of rough joints, through both test results and published data.</p>