Attributes of several thousand fractures were collected in three boreholes of 2.2, 3.6, and 3.8 km depth, penetrating the Soultz Hot Dry Rock reservoir (France). The fractures were sampled from cores and from several high-resolution imaging techniques such as borehole televiewer (BHTV), ultrasonic borehole imager (UBI), formation microscanner (FMS), formation microimager (FMI), and azimuthal resistivity imaging (ARI). A comparison was made between the data collected on cores and those provided by different imaging techniques. The comparison clearly establishes that the different wall-images are not as exhaustive as the core data and cannot provide a complete characterization of the fracture network. Discrete fractures thinner than 1 mm are not properly detected. This is also the case for discrete fractures closer than 5 mm, which appear only as single traces. The imaging techniques are, nevertheless, very powerful for characterizing altered fracture clusters. Whatever the technique used, the fracture strikes were correctly sampled with the different systems. This comparison allowed us to calibrate the fracture population data obtained from the imaging system in order to correct for the filtering effect introduced by the technique itself and by the alteration of the rock mass.¿ 1997 American Geophysical Union