oa Reducing Noise in 3D Seismic Data While Preserving Structural Detail

We present a new data adaptive method for smoothing 3D post-stacked seismic attributes. The<br>method can reduce random noise while preserving the structure without prior computation of the<br>structure orientation. It works as follows: within a neighborhood sub-window, we smooth the data<br>along a set of pre-defined orientations; the best result is then selected as output. This best orientation<br>often approximates to the true structure orientation embedded in the data; therefore, the embedded<br>structure is preserved. The selection rule for the “best” orientation depends on the data type and<br>purpose of maneuver; it can be minimum deviation, and maximum, minimum or absolute-maximum<br>summation. The scheme can be further combined with median, alpha-trim, symmetric near neighbor,<br>or edge preserving filters. A stop mechanism can be built-in when best orientations cannot be<br>determined. Our results show that it is an effective way to reduce random noise, eliminate footprint<br>and to enhance coherence and curvature attributes. It can also be applied to seismic amplitude to<br>enhance auto-picking of horizons, first arrivals, or refraction events.