Processing 3-D georadar data acquired across rugged terrains is a challenging and time-consuming task. We have recorded 3-D georadar data and corresponding coordinate information on an unstable mountainside in the Swiss Alps. The primary objective was to detect active fracture zones. After passing the data through a conventional processing scheme (i.e. dewow filtering, time-zero alignment, gridding, amplitude scaling, low-pass frequency filtering, and conventional topographic static corrections), the resultant 3-D images were plagued by artificial discontinuities (e.g. anomalous time shifts and signal variations) and data gaps caused by coordinate inaccuracies and radiation pattern distortions associated with the presence of abrupt topographic variations and large boulders. We resolved these problems by passing the data through an adaptive f-xy deconvolution routine followed by f-kx and f-ky filtering. This scheme also reduced incoherent noise. Finally, the data were migrated using a 3-D algorithm that fully accounted for the rugged topography. In the migrated images, a moderately dipping reflector and a suite of prominent point diffractors could be projected to open fracture zones at the surface.