Diffraction Separation Using the Crs Technique: A Field Data Application

Diffracted waves are often associated with geological structures like faults, pinchouts, wedgeouts or a sudden change in facies (Kanasewich and Phadke, 1988). Identification of such structures in a seismic or ground penetrating radar (GPR) image is highly<br>dependent on our ability to utilize the diffracted energy. Unfortunately, diffractions often manifest themselves on seismic (or GPR) data with a much weaker signal strength compared to reflections and they often fall within the noise level. As a consequence, classical signal processing methods treat diffractions as noise and imaging is carried out in favor of reflections. Recently, however, different approaches have been proposed to separate diffractions from reflections so that additional high-resolution information can be obtained from direct imaging of the diffracted energy. In this paper, we propose to perform diffraction and reflection separation based on the Common Reflection Surface (CRS) concept. Within this formulation, suppression (or attenuation) of reflections is carried out by selecting the appropriate stacking surface for diffractions based on a coherency measure. Here we tested both Semblance and MUltiple SIgnal Classification (MUSIC) as a coherency measure for the CRS parameter estimation. The potential application of the technique has been demonstrated employing a multi-offset GPR dataset.