This paper analyzes the performance of several seismic data denoising algorithms, measured in the extent of noise suppression and the level of unintended signal distortion. It shows that the error in the filtered signal consists of residual noise, and signal distortion due to model-order mismatch and subspace corruption by noise. The contributing effects of these components to the overall denoising performance depend on algorithm parameters, input signal versus noise ratio (SNR), as well as the general signal and noise structure. The analysis focuses on singular spectrum analysis (SSA) and its robust version, and f-x deconvolution. However extension to other denoising algorithms is possible. While the main goal is to improve understanding of denoising performance and behavior, the analysis also hopes to provide a basis for adaptive adjustment of denoising parameters, paving the way for automated processing of large-scale field data where signal and noise structure and SNR level can vary significantly from window to window, or cube to cube.
Presentation Date: Wednesday, October 19, 2016
Start Time: 8:50:00 AM
Location: 148
Presentation Type: ORAL
