This short course introduces the core concepts of predictability, sparsity, and low-rank structure that underpin modern seismic data-conditioning algorithms.
Participants will learn how these principles enhance signal-to-noise ratio, stabilize traces, and recover missing or corrupted data. Through practical examples, we demonstrate how classical tools—prediction filters, sparse deconvolution, and SSA/Cadzow filtering—can be applied to real seismic traces.
The course builds intuition that naturally extends to multidimensional methods such as F–X prediction, MSSA, and 5D interpolation/Compressive Solutions. By the end, attendees gain a clear understanding of why predictability, sparsity, and low-rank structure remain fundamental to robust seismic data processing and algorithm development.
Duration
One Day
Intended Audience
- Seismic data processing contractors
- R&D geophysicists
- Seismic data-processing professionals
- Graduate students with an interest in data preconditioning and algorithm development
Level
Intermediate and advanced
Prerequisites (Knowledge/Experience/Education required)
BSc Geophysics, Math, Physics, etc., and programming. Assume attendees are comfortable with Linear Algebra and Fourier Analysis.
Learner Outcomes
By the end of this short course, participants will be able to:
- Construct sparsity-promoting template algorithms and apply them to problems such as Radon transforms, 5D Fourier reconstruction, and seismic deblending. Examine new paradigms of Compressive Sensing with examples.
- Develop and implement prediction-error and projection-based filters for SNR enhancement and distinguish the key predictability principles underlying methods such as Spitz F–X trace reconstruction for up-sampling aliased data.
- Classify and apply reduced-rank filtering techniques—including SSA/Cadzow and related methods—for random-noise attenuation, interpolation, and coherent-noise suppression, and interpret recent advances in tensor-based reconstruction.
- Design and code practical signal-processing workflows that can serve as foundational modules for larger seismic data-processing platforms.
- Discuss how classical signal-processing concepts can be integrated with modern ML/AI-based approaches
