Imaging Contrasted Media with Total Variation Constrained Full Waveform Inversion and Split Bregman Iterations
H. Aghamiry, A. Gholami and S. Operto
Event name: 80th EAGE Conference and Exhibition 2018
Session: FWI II - Cycle Skipping And Constrained FWI
Publication date: 11 June 2018
Info: Extended abstract, PDF ( 1.09Mb )
Price: € 20
Full waveform inversion (FWI) is a waveform matching procedure which provides subsurface model with a resolution of the order of the seismic wavelength. However, this high resolution potential also makes the inverse problem underlying FWI highly nonlinear and ill-posed and highlights the role of regularization. To stabilize FWI while preserving its resolution power, frequency-domain FWI is recasted as a constrained total-variation (TV) minimization of the subsurface parameters where the inequality constraints require the modelled wavefields to fit the data and satisfy the wave equation with prescribed errors. The wavefield and subsurface parameters are updated through an alternating optimization technique based on split Bregman iterations. At each iteration of the optimization, the wavefield is updated via a penalty method to best fit the data and satisfy the wave equation, while the constrained total-variation optimization provides subsurface parameter updates which conciliate blockiness and good signal-to-noise ratio. The resilience of the method to noise and cycle skipping as well as its resolution power are illustrated with two synthetic examples performed with a toy example and the 2004 Bp salt model.Starting from crude initiam model, we show how the proposed method manages sharp velocity contrasts generated by salt bodies and improves sub-salt imaging accordingly.