Ground Penetrating Radar (GPR) (or UWB-SP Ultra Wide Band Short Pulse) radar are widely used in many industrial and military applications for the detection of buried objects and more recently for detecting antipersonnel mines with low metallic content. In this work we report a new algorithm for extracting the characteristic response of buried mines by processing three-dimensional data sets obtained with a scanning probe placed at the air-ground interface. The new algorithm requires few assumptions concerning the experimental conditions. The input parameters of the program are the spatial and time sampling steps and a rough estimate of the propagation velocity in the soil. Moreover we introduce the concept of local echo contrast pattern for non point-like objects in order to overcome the limitations of standard signal processing approaches based on the assumption of a hyperbolic response from a point-like reflector. The information on the mine shape is derived by the time-flight measurements performed on each contrast pattern. The method is capable of retrieving a series of mine contours and an associated propagation velocity. A pictorial visualization technique in 3-d space is proposed showing the estimated contours and the depth from the top surface and compared to standard C-scan visualization.