GNGTS 2013 - Atti del 32° Convegno Nazionale

concave event. The analytical expressions of reflection and diffraction moveouts in constant velocity media can be found in Landa et al. (2008). In particular, in the zero offset case, the imprint of a flat reflector in the dip-angle gather located at x is the following: Here is the medium velocity, is the migration velocity, is the dip angle, is the actual reflector’s dip. The reflector is described by the equation: . Note that, if the migration is performed with the correct velocity ( ), then the stationary point occurs at the actual reflector’s dip: . The shape of a diffractor point in the dip-angle gather located at x is: where . Thus, when migrating with the correct velocity, the shape of a diffraction point in the gather located exactly above is a horizontal flat line which corresponds to illuminating the diffractor uniformly from all directions. If the gather is not located just above the diffractor the shape is a curve with no stationary points. It is possible to take advantage of the local discrepancy between reflection and diffraction events and use dip-filtering in this domain, in order to separate diffractions which do not follow the dominant trend of dips. Being substantially a pre-stack domain, the dip-angle gathers are more prone to noise with respect to the post-stack image domain and this can lead to an incorrect dip estimation which causes artifacts in the dip-filtered gather and in the subsequent migrated image of the diffractors (i. e. the stack over the dip angles). On the other hand the dip-angles domain has the advantage that the shape of diffractions is clearly identified and other kind of spurious events cannot be mistaken for them. Diffractions look significantly different from reflections in the common-image gathers and unlike reflections, they are affected by velocity errors in the conventional way. These two characteristics mean that also this domain could be used to perform effective migration velocity analysis by using the moveout of diffraction events. Combined separation technique. The aim of our work was to unify the two methodologies mentioned above in order to exploit the peculiarity of each. The input data is depth migrated image in the dip-angle CIG domain. The algorithm is made of the three following steps: 1. local slopes are estimated on the stack image through a filter based on the gradient square tensor method; 2. dip filtering of each constant dip angle image is performed. It removes strong coherent events with continuously variable slopes calculated at step 1. In this way, being the filter a linear operator, we obtain the contribution of each common angle panel to the final residual consisting in all events, including seismic diffractions, that do not follow the dominant slope pattern. After this step we have eliminated the reflection but, besides the diffractions, there are still spurious events due to migration artifacts, uneven illumination or noise; 3. each dip-angle CIG is then further processed, this time in a way that emphasizes the coherent events through the local coherency of the filter. Indeed, in the dip angle domain, the diffracted events are strong coherent events (horizontal events if migrated with the correct velocity), on the other hand the remaining spurious events do not usually show a clearly identifiable dip. Therefore by applying a mask proportional to the local coherency of the dips estimated in the dip angle CIGs the true diffractions are emphasized compared with the spurious events. 73 GNGTS 2013 S essione 3.1