Ground roll removal using adaptive reflection mask function in Curvelet domain

Ground roll is one of the most common coherent noise produced by surface waves during land seismic data acquisition. The removal of ground roll is of great importance during preprocessing, as it can conceal the smallamplitude reflected events. Curvelet transform decomposes
seismic events in a time/space window based on their local frequency and dip information. Thus, the separation of ground roll noise and reflected events can be accomplished
better in the curvelet domain than in other transformations. In this paper, we present a new curvelet-based ground roll suppression method. First, curvelet coefficient mask functions for reflection events are generated based on adaptive threshold for fine-scale panels. The reflection mask function will be downscaled to all scales and dipping angles. We then calculate another ground roll mask function in coarse scales. Combining these two mask functions we can generate the final adaptive reflection mask function for all scales and dipping angles. The adaptive mask function will be used to mute out curvelet coefficient and inverse transform to get the ground roll attenuated image. The real data tests indicate the proposed method can effectively suppress the ground roll energy and reveal the hidden reflection signals.