Seismic Noise Attenuation Using Curvelet Transform and Dip Map Data Structure

The curvelet transform is a known tool used in the attenuation of coherent and incoherent noise in seismic data. It utilizes the fact that signal and noise are usually better separated in the curvelet domain than in the time-space (TX) domain. Coefficients of the transform are not independent and neighboring coefficients are strongly correlated, which existing curvelet-based noise attenuation algorithms do not fully utilize. In this work, we propose to use a data structure called a dip map to describe dip information in seismic data. This information links local curvelet coefficients together in adaptive thresholding or subtraction of curvelet coefficients in seismic denoising algorithms. We used the dip map to improve the curvelet multiple subtraction algorithm and the results show significant improvement over traditional methods with real data.