Time-frequency representation for seismic data using sparse S transform

Abstract

The S transform is a time-frequency representation with multi-scale focus. It adopts a scalable Gaussian window function to provide a frequency dependent resolution. However, it still suffers from low resolution, which does not satisfy the high precision seismic imaging. Therefore, we propose the sparse S transform to obtain a sparse and aggregated time-frequency spectrum, and apply it into seismic data analysis. The S transform is considered as inverse problem with L1 minimization constraint known as basis pursuit denoising (BPDN) form. The good performance of the proposed method is assessed on simulated and real seismic data. The results indicate that our method can enhance the sparsity of ST, and provide a high resolution and focused time-frequency spectrum for seismic data, which is conducive to seismic imaging and reservoir interpretation.