Denoising of gravitational-wave signal GW150914 via total-variation methods

TL;DR

This paper demonstrates that total variation (TV) denoising methods effectively remove non-Gaussian, non-stationary noise from gravitational wave signals like GW150914, providing an alternative to Bayesian approaches for waveform reconstruction.

Contribution

The study extends TV-denoising techniques to real gravitational wave data, showing they can match Bayesian methods in waveform quality for GW150914.

Findings

TV-denoising successfully removes complex noise from GW150914

Waveform quality comparable to Bayesian methods

Applicable to real non-Gaussian, non-stationary noise

Abstract

We apply a regularized Rudin-Osher-Fatemi total variation (TV) method to denoise the transient gravitational wave signal GW150914. We have previously applied TV techniques to denoise numerically generated grav- itational waves embedded in additive Gaussian noise, obtaining satisfactory results irrespective of the signal morphology or astrophysical origin. We find that the non-Gaussian, non-stationary noise from the gravitational wave event GW150914 can also be successfully removed with TV-denoising methods. The quality of the de- noised waveform is comparable to that obtained with the Bayesian approach used in the discovery paper [1]. TV-denoising techniques may thus offer an additional viable approach for waveform reconstruction.