In this study we propose a novel methodology to characterize the roughness of rock joints using photogrammetry techniques and frequency analyses. The process starts with the generation of a 3D model (a dense point cloud) using the Structure from Motion (SfM) technology. Joint profiles are then extracted from this model and their frequency spectra are obtained using the Fast Fourier Transform. Finally, several approaches to parametrize the amplitude-frequency relationships are proposed, so that the roughness of the joint profiles are characterized with the fitting parameters. The methodology can differentiate between waviness and roughness, so that it could be used in future analysis of the shear behaviour of joints affected by this distinction. The proposed methodology is applied to two samples of granite representative of an unweathered fresh-cut surface with natural roughness. The results from the parametrization of the frequency spectra are used to carry out a classification analysis to study if their best-fitting parameters can estimate the Joint Roughness Coefficient (JRC) adequately. Results show that JRC values estimated with the spectral information tend to be quite similar (with errors less or equal to ±2 in about 80% of cases) to those estimated using the Ζ 2 statistical parameter, therefore validating the use of frequency spectra to characterize the roughness of rock joints.