Model-Fitting Clustering for Harmonic Noise Attenuation

Summary In Vibroseis data, source-generated harmonic distortions are a frequently-observed distinctive phenomenon and are typically regarded as unwanted noise. An inversion-based method to separate the fundamental and higher order harmonics from distorted sweeps in the Gabor (time-frequency) domain was recently proposed and successfully applied to field data. However, after correlating the separated harmonics with the field data, lower-harmonics-related artifacts will remain and hinder the imaging. In this study, we propose to utilize the RANdom SAmple Consensus algorithm (RANSAC) for designing appropriate polygonal filters to attenuate this type of artifacts and enable high resolution broadband seismic imaging utilizing harmonics signals.