Image synthesis with graph cuts: a fast model proposal mechanism in probabilistic inversion

Zahner, Tobias; Lochbühler, Tobias; Mariethoz, Grégoire; Linde, Niklas

doi:10.1093/gji/ggv517

Cited by 45 publications

(48 citation statements)

References 63 publications

(79 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our first 2‐D test case considers the 2,500 × 2,500 binary channelized TI depicted in Figure a. This TI is a hand‐made drawing inspired by Strebelle's TI (Strebelle, ) by Zahner et al (). For learning, we used

z_{normalx} = 12

and q = 1.…”

Section: Geostatistical Simulation Resultsmentioning

confidence: 99%

“… Three training images (TIs) used for the considered geostatistical simulation tests. The TIs are (a) the 2,500 × 2,500 hand‐made drawing inspired by Strebelle's TI by Zahner et al (), (b) the 400 × 400 braided river aquifer by Pirot et al () and (c) and (d) the

180 \times 150 \times 120

categorical fold aquifer from http://www.trainingimages.org/training-images-library.html. Subfigures (c) and (d) present two different views of the same TI.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Training‐Image Based Geostatistical Inversion Using a Spatial Generative Adversarial Neural Network

Laloy

Hérault

Jacques

et al. 2018

Water Resources Research

Self Cite

310

284

View full text Add to dashboard Cite

Probabilistic inversion within a multiple‐point statistics framework is often computationally prohibitive for high‐dimensional problems. To partly address this, we introduce and evaluate a new training‐image based inversion approach for complex geologic media. Our approach relies on a deep neural network of the generative adversarial network (GAN) type. After training using a training image (TI), our proposed spatial GAN (SGAN) can quickly generate 2‐D and 3‐D unconditional realizations. A key characteristic of our SGAN is that it defines a (very) low‐dimensional parameterization, thereby allowing for efficient probabilistic inversion using state‐of‐the‐art Markov chain Monte Carlo (MCMC) methods. In addition, available direct conditioning data can be incorporated within the inversion. Several 2‐D and 3‐D categorical TIs are first used to analyze the performance of our SGAN for unconditional geostatistical simulation. Training our deep network can take several hours. After training, realizations containing a few millions of pixels/voxels can be produced in a matter of seconds. This makes it especially useful for simulating many thousands of realizations (e.g., for MCMC inversion) as the relative cost of the training per realization diminishes with the considered number of realizations. Synthetic inversion case studies involving 2‐D steady state flow and 3‐D transient hydraulic tomography with and without direct conditioning data are used to illustrate the effectiveness of our proposed SGAN‐based inversion. For the 2‐D case, the inversion rapidly explores the posterior model distribution. For the 3‐D case, the inversion recovers model realizations that fit the data close to the target level and visually resemble the true model well.

show abstract

z_{normalx} = 12

and q = 1.…”

Section: Geostatistical Simulation Resultsmentioning

confidence: 99%

180 \times 150 \times 120

categorical fold aquifer from http://www.trainingimages.org/training-images-library.html. Subfigures (c) and (d) present two different views of the same TI.…”

Section: Methodsmentioning

confidence: 99%

Training‐Image Based Geostatistical Inversion Using a Spatial Generative Adversarial Neural Network

Laloy

Hérault

Jacques

et al. 2018

Water Resources Research

Self Cite

310

284

View full text Add to dashboard Cite

show abstract

“…Just like SGS, multiple-point simulations could replace eigen and Cholesky decompositions in InSAR workflows. It also shows promising applications in a Monte Carlo setting (Pirot et al, 2017;Zahner et al, 2015). Tweaking its parameters helps balance speed and quality of the results (Meerschman et al, 2013), and the DS can be used in an optimization or Monte Carlo setting (Mariethoz et al, 2010b), which could be of major use in InSAR processing (e.g., Elliott et al, 2010;Wright et al, 2003).…”

Section: Discussionmentioning

confidence: 99%

“…One solution stands out: simulating patches of cells instead of a single cell at a time (e.g., Hoffimann et al, 2017;Li et al, 2016;Tahmasebi & Sahimi, 2016). It also shows promising applications in a Monte Carlo setting (Pirot et al, 2017;Zahner et al, 2015).…”

Section: Discussionmentioning

confidence: 99%

Generative Modeling of InSAR Interferograms

Rongier

Rude

Herring

et al. 2019

Earth and Space Science

View full text Add to dashboard Cite

Interferometric synthetic aperture radar (InSAR) has become an essential technique to detect surface variations due to volcanoes, earthquakes, landslides, glaciers, and aquifers. However, Earth's ionosphere, atmosphere, vegetation, surface runoff, etc., introduce noise that requires post-processing to separate its components. This work defines a generator to create interferograms that include each of those components. Our approach leverages deformation models with real data, either directly or through machine learning using geostatistical methods. These methods result from previous developments to more efficiently and better simulate spatial variables and could replace some statistical approaches used in InSAR processing. We illustrate the use of the generator to simulate an artificial interferogram based on the 2015 Illapel earthquake and discuss the improved performance offered by geostatistical approaches compared with classical statistical ones. The generator establishes a tool for multiple applications (1) to evaluate InSAR correction workflows in controlled scenarios with known ground truth; (2) to develop training sets and generative methods for machine learning algorithms; and (3) to educate on InSAR and its principles.

show abstract

“…Others include mutual information (Mandolesi and Jones 2014), or geostatistical methods in a probabilistic inversion (e.g., Zahner et al 2016). Bosch (2016) and Hansen et al (2016) discuss the theoretical basis of using geological prior information in more detail.…”

Section: Cooperative and Constrained Inversionmentioning

confidence: 99%

Integrating Electromagnetic Data with Other Geophysical Observations for Enhanced Imaging of the Earth: A Tutorial and Review

Moorkamp

2017

Surv Geophys

View full text Add to dashboard Cite

In this review, I discuss the basic principles of joint inversion and constrained inversion approaches and show a few instructive examples of applications of these approaches in the literature. Starting with some basic definitions of the terms joint inversion and constrained inversion, I use a simple three-layered model as a tutorial example that demonstrates the general properties of joint inversion with different coupling methods. In particular, I investigate to which extent combining different geophysical methods can restrict the set of acceptable models and under which circumstances the results can be biased. Some ideas on how to identify such biased results and how negative results can be interpreted conclude the tutorial part. The case studies in the second part have been selected to highlight specific issues such as choosing an appropriate parameter relationship to couple seismic and electromagnetic data and demonstrate the most commonly used approaches, e.g., the cross-gradient constraint and direct parameter coupling. Throughout the discussion, I try to identify topics for future work. Overall, it appears that integrating electromagnetic data with other observations has reached a level of maturity and is starting to move away from fundamental proof-of-concept studies to answering questions about the structure of the subsurface. With a wide selection of coupling methods suited to different geological scenarios, integrated approaches can be applied on all scales and have the potential to deliver new answers to important geological questions.

show abstract

Image synthesis with graph cuts: a fast model proposal mechanism in probabilistic inversion

Cited by 45 publications

References 63 publications

Training‐Image Based Geostatistical Inversion Using a Spatial Generative Adversarial Neural Network

Training‐Image Based Geostatistical Inversion Using a Spatial Generative Adversarial Neural Network

Generative Modeling of InSAR Interferograms

Integrating Electromagnetic Data with Other Geophysical Observations for Enhanced Imaging of the Earth: A Tutorial and Review

Contact Info

Product

Resources

About