Frequency-domain rank-reduction applications in seismic processing
Presenter: Stewart Trickett
Presented at SEG 2015 as a post-convention workshop
Friday, October 23, 8:30 AM – Noon
Ernest L. Morial Convention Center, Room 215/216
New Orleans, LA, USA
Summary: Over the last 15 years, matrix rank reduction methods have been developed for noise removal and interpolation of seismic data. These methods have primarily been applied on constant-frequency slices, as a key theorem states that the signal should have low rank in this domain. Over the years these filters have been extended to multiple dimensions, robust filtering for erratic noise, source deblending, de-aliasing, tensor rank reduction, coherent noise removal, automatic rank determination, and computational speedups. I will briefly describe these developments and show many examples. I will also list some open questions and avenues for future research.