Natural fractures are characterized by rough surfaces and complex fluid flows. A large distribution of apertures (residual voids) within their walls and the presence of contact points produce heterogeneous flows (channelling). The resulting permeabilities, porosities or fluid-rock exchange surfaces cannot be realistically modelled by parallel and smooth plate models. It is therefore very important to better constrain models of the fracture planes : asperity heights and aperture distribution to be able to fix specific models of permeability in specific fracture void geometry. In this approach, a precise description of the fracture surface planes is given by providing some new quantitative data of surface roughness in the case of natural fractures. Studied fractures are sampled in a granite and a sandstone in the deep basement of the Hot Dry Rock site of Soultzsous-Forêts (Bas-Rhin, France). An original use of close-range photogrammetry is performed to quantify XYZ data on fracture walls. This methodology is presented as a non destructive, precise and accurate technology to quantify some digital terrain models (DTM) of the fracture plane topography. XYZ results are statistically treated in terms of surface roughness and tortuosity and are compared for different rocks to previous data obtained by mechanical profilometry. The results shows that the photogrammetric approach gives same order of asperity heights magnitudes as profilometry despite a shift towards more important values of roughness when close range photogrammetry uses relative autocorrelation models. The advantage of photogrammetry is that this technique gives very quick results and is non destructive when thin alteration or pulverulent deposits are present within fracture walls or in the rock matrix. The disadvantage is that a slight smoothing of data is inherent to an absolute model calibration. Finally only relative 300*300 DTM are finally chosen to match profilometry data because of their higher precision in terms of micro roughness description to compare natural fracture surfaces. In the objective of a classification of fracture roughness in specific geological contexts, the photogrammetric approach gives a good estimation of different classes of roughness in function of rock alteration and type.