Building a regularization toolbox for FWI

A versatile procedure has been developed in the framework to ADMM-based WRI to interface different regularizations with the optimization algorithm where WRI is broken down into two successive linear subproblems. We adapt this procedure to regularize classical nonlinear FWI. This procedure is reviewed in the WIND Technical Report 1 for a vast catalogue of regularizer. A regularization toolbox based upon this approach is under development.

Aghamiry H., A. Gholami and S. Operto (2020). A practical guide to implement smooth and nonsmooth regularization in Full Waveform Inversion with ADMM, Technical Report n 1 - version 1, WIND project, 29 pages .

fig_bp_regy.png

Figure 1: Performance of different regularizer in IR-WRI for the 2004 BP salt model. (a) True velocity model/ (b) Starting velocity model. (c) Zero-order Tikhonov regularization. (d) TV regularization. (e) Sedond-order Tikhonov regularization. (f) Compund Tikhonov + TV regularization.

More details can be found in H. Aghamiry, A. Gholami and S. Operto. Compound Regularization of Full-waveform Inversion for Imaging Piecewise Media, IEEE Transactions on Geoscience and Remote Sensing, https://doi.org/10.1109/TGRS.2019.2944464, 58(2), pages 1192--1204, 2020.
Topic attachments
I Attachment Action Size Date Who Comment
fig_bp_regy.pngpng fig_bp_regy.png manage 815.7 K 08 May 2020 - 08:06 Main.UnknownUser  
fig_denoising_l1_bp.pngpng fig_denoising_l1_bp.png manage 2183.5 K 08 May 2020 - 08:03 Main.UnknownUser