The Ground Probing Radar (GPR) is a valuable tool for near surface geological, geotechnical, engineering, environmental, archaeological and other work. GPR images of the subsurface frequently contain geometric information (constant or variable-dip reflections) from various structures such as bedding, cracks, fractures etc. Such features are frequently the target of the survey; however, they are usually not good reflectors and they are highly localized in time and in space. Their scale is therefore a factor significantly affecting their detectability. At the same time, the GPR method is very sensitive to broadband noise from buried small objects, electromagnetic anthropogenic activity and systemic factors, which frequently blurs the reflections from such targets. The purpose of this paper is to investigate the Curvelet Transform (CT) as a means of S/N enhancement and information retrieval from 2-D GPR sections, with particular emphasis on the recovery of features associated with specific temporal or spatial scales and geometry (orientation/dip).The CT is a multiscale and multidirectional expansion that formulates an optimally sparse representation of bivariate functions with singularities on twice-differentiable (C2-continuous) curves (e.g. edges) and allows for the optimal, whole or partial reconstruction of such objects. The CT can be viewed as a higher dimensional extension of the wavelet transform: whereas discrete wavelets are isotropic and provide sparse representations of functions with point singularities, curvelets are highly anisotropic and provide sparse representations of functions with singularities on curves. A GPR section essentially comprises a spatio-temporal sampling of the transient wavefield which contains different arrivals that correspond to different interactions with wave scatterers in the subsurface (wavefronts). These are generally longitudinally piecewise smooth and transversely oscillatory, i.e. they comprise edges. Curvelets can detect wavefronts at different angles and scales because curvelets of a given angle and scale locally correlate with aligned wavefronts of the same scale.The utility of the CT in processing noisy GPR data is investigated with software based on the Fast Discrete CT and adapted for use with a set of interactive driver functions that compute and display the curvelet decomposition and then allow the manipulation of data (wavefront) components at different scales and angles via the corresponding manipulation (cancellation or restoration) of their associated curvelets. The method is demonstrated with data from archaeometric, geotechnical and hydrogeological surveys, contaminated by high levels of noise, or featuring straight and curved reflections in complex propagation media, or both. It is shown that the CT is very effective in enhancing the S/N ratio by isolating and cancelling directional noise wavefronts of any scale and angle of emergence, sometimes with surgical precision and with particular reference to clutter. It can as successfully be used to retrieve waveforms of specific scale and geometry for further scrutiny, also with surgical precision, as for instance distinguish signals from small and large aperture fractures and faults, different phases of fracturing and faulting, bedding etc. Moreover, it can be useful in investigating the characteristics of signal propagation (hence material properties), albeit indirectly. This is possible because signal attenuation and temporal localization are closely associated, so that scale and spatio-temporal localization are also closely related. Thus, interfaces embedded in low attenuation domains will tend to produce sharp reflections and fine-scale localization. Conversely, interfaces in high attenuation domains will tend to produce dull reflections with broad localization.