The 10 standard roughness joint profiles provided a visual comparison to get the joint roughness coefficient (JRC) of rock joint surface, but the accuracy of this method is influenced by human factors. Therefore, many researchers try to evaluate the roughness morphology of joint surface through the statistical parameter method. However, JRC obtained from most of the existing statistical parameters did not reflect the directional property of joint surface. Considering the 10 standard profiles as models of different roughness joints, we proposed a new idea for the accurate estimation of JRC. Based on the concept of area difference, the average of positive area difference (Sa) and sum of positive area difference (Ss) were first proposed to reflect the roughness of joint surfaces on the basis of directional property, and their fitting relationship with JRC was also investigated. The result showed that the Sa and Ss calculated by shearing from right to left (FRTL) and JRC backcalculated from right to left (FRTL) came to a satisfying power law. The correlation between JRC and Sa was better than that of Ss. The deviation between the predicted value calculated by Sa and the true value was smaller than that obtained from the existing statistical parameters. Therefore, Sa was recommended as a new statistical parameter to predict the JRC value of joint profile. As the sampling interval increased from 0.5 to 4 mm, the correlation between Sa and JRC gradually decreased, and the accuracy of the prediction results also declined. Compared with the single JRC values for joint profiles mentioned in the literature, the forward and reverse JRC were obtained. Based on the laboratory direct shear test of the natural joint surface, the JRC values of two joint surfaces in four shear directions were backcalculated by the JRC-JCS model. Based on 3D scanning and point cloud data processing technology, JRC of joint surface in different directions were obtained by Sa method, and they are very close to those obtained by JRC-JCS model. It is confirmed that Sa could accurately estimate the joint roughness coefficient and reflect its anisotropy.