Automated {RGB} {Frequency} {Selection} for {Spectral} {Decomposition} via {Sparse} {Picks} and 2D {Gaussian} {Fitting}.

D. Henaux and Laurent Souche and N. Daynac and H. Abdallah and M. Papin and Q. Corlay. ( 2026 )
in: 2026 {RING} meeting, pages 185--187, ASGA

Abstract

Spectral decomposition is widely used to highlight stratigraphic features and resolve bedset thickness (Partyka et al., 1999). Optimal frequency selection for RGB rendering is typically performed by manual trial-and-error on spectrograms, which is a time-consuming, and operator-dependent process. We present a methodology that automates this selection based on sparse interpreter picks on stratal slices. Seismic traces of interest are automatically selected by similarity, aligned, and stacked. Then, a 2D Gaussian function is fitted to the resulting time-frequency spectrogram via PCA clustering, and the three RGB frequencies are derived directly from the best-fit Gaussian parameters. This methodology improves reproducibility, reduces subjectivity, and accelerates the workflow. Finally, we also outline a sparsification approach to produce cleaner spectrograms as input to the clustering step.

Download / Links

BibTeX Reference

@inproceedings{Henaux2026RM,
 abstract = {Spectral decomposition is widely used to highlight stratigraphic features and resolve bedset thickness (Partyka et al., 1999). Optimal frequency selection for RGB rendering is typically performed by manual trial-and-error on spectrograms, which is a time-consuming, and operator-dependent process. We present a methodology that automates this selection based on sparse interpreter picks on stratal slices. Seismic traces of interest are automatically selected by similarity, aligned, and stacked. Then, a 2D Gaussian function is fitted to the resulting time-frequency spectrogram via PCA clustering, and the three RGB frequencies are derived directly from the best-fit Gaussian parameters. This methodology improves reproducibility, reduces subjectivity, and accelerates the workflow. Finally, we also outline a sparsification approach to produce cleaner spectrograms as input to the clustering step.},
 author = {Henaux, D. and Souche, L. and Daynac, N. and Abdallah, H. and Papin, M. and Corlay, Q.},
 booktitle = {2026 {RING} meeting},
 language = {en},
 pages = {185--187},
 publisher = {ASGA},
 title = {Automated {RGB} {Frequency} {Selection} for {Spectral} {Decomposition} via {Sparse} {Picks} and 2D {Gaussian} {Fitting}.},
 year = {2026}
}