A comparative study of structural similarity and regularization for joint inverse problems governed by PDEs

Benjamin Crestel, Georg Stadler, Omar Ghattas

Research output: Contribution to journalArticle

Abstract

Joint inversion refers to the simultaneous inference of multiple parameter fields from observations of systems governed by single or multiple forward models. In many cases these parameter fields reflect different attributes of a single medium and are thus spatially correlated or structurally similar. By imposing prior information on their spatial correlations via a joint regularization term, we seek to improve the reconstruction of the parameter fields relative to inversion for each field independently. One of the main challenges is to devise a joint regularization functional that conveys the spatial correlations or structural similarity between the fields while at the same time permitting scalable and efficient solvers for the joint inverse problem. We describe several joint regularizations that are motivated by these goals: a cross-gradient and a normalized cross-gradient structural similarity term, the vectorial total variation, and a joint regularization based on the nuclear norm of the gradients. Based on numerical results from three classes of inverse problems with piecewise-homogeneous parameter fields, we conclude that the vectorial total variation functional is preferable to the other methods considered. Besides resulting in good reconstructions in all experiments, it allows for scalable, efficient solvers for joint inverse problems governed by PDE forward models.

Original languageEnglish (US)
Article number024003
JournalInverse Problems
Volume35
Issue number2
DOIs
StatePublished - Feb 1 2019

Fingerprint

Structural Similarity
Inverse problems
Comparative Study
Regularization
Inverse Problem
Spatial Correlation
Total Variation
Gradient
Inversion
Simultaneous Inference
Prior Information
Term
Experiments
Attribute
Norm
Numerical Results
Model
Experiment

Keywords

  • cross-gradient
  • joint inversion
  • multi-physics inverse problem
  • nuclear norm
  • structural similarity prior
  • vectorial total variation

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Signal Processing
  • Mathematical Physics
  • Computer Science Applications
  • Applied Mathematics

Cite this

A comparative study of structural similarity and regularization for joint inverse problems governed by PDEs. / Crestel, Benjamin; Stadler, Georg; Ghattas, Omar.

In: Inverse Problems, Vol. 35, No. 2, 024003, 01.02.2019.

Research output: Contribution to journalArticle

@article{71a805b527d74816aa34dd8e834a4225,
title = "A comparative study of structural similarity and regularization for joint inverse problems governed by PDEs",
abstract = "Joint inversion refers to the simultaneous inference of multiple parameter fields from observations of systems governed by single or multiple forward models. In many cases these parameter fields reflect different attributes of a single medium and are thus spatially correlated or structurally similar. By imposing prior information on their spatial correlations via a joint regularization term, we seek to improve the reconstruction of the parameter fields relative to inversion for each field independently. One of the main challenges is to devise a joint regularization functional that conveys the spatial correlations or structural similarity between the fields while at the same time permitting scalable and efficient solvers for the joint inverse problem. We describe several joint regularizations that are motivated by these goals: a cross-gradient and a normalized cross-gradient structural similarity term, the vectorial total variation, and a joint regularization based on the nuclear norm of the gradients. Based on numerical results from three classes of inverse problems with piecewise-homogeneous parameter fields, we conclude that the vectorial total variation functional is preferable to the other methods considered. Besides resulting in good reconstructions in all experiments, it allows for scalable, efficient solvers for joint inverse problems governed by PDE forward models.",
keywords = "cross-gradient, joint inversion, multi-physics inverse problem, nuclear norm, structural similarity prior, vectorial total variation",
author = "Benjamin Crestel and Georg Stadler and Omar Ghattas",
year = "2019",
month = "2",
day = "1",
doi = "10.1088/1361-6420/aaf129",
language = "English (US)",
volume = "35",
journal = "Inverse Problems",
issn = "0266-5611",
publisher = "IOP Publishing Ltd.",
number = "2",

}

TY - JOUR

T1 - A comparative study of structural similarity and regularization for joint inverse problems governed by PDEs

AU - Crestel, Benjamin

AU - Stadler, Georg

AU - Ghattas, Omar

PY - 2019/2/1

Y1 - 2019/2/1

N2 - Joint inversion refers to the simultaneous inference of multiple parameter fields from observations of systems governed by single or multiple forward models. In many cases these parameter fields reflect different attributes of a single medium and are thus spatially correlated or structurally similar. By imposing prior information on their spatial correlations via a joint regularization term, we seek to improve the reconstruction of the parameter fields relative to inversion for each field independently. One of the main challenges is to devise a joint regularization functional that conveys the spatial correlations or structural similarity between the fields while at the same time permitting scalable and efficient solvers for the joint inverse problem. We describe several joint regularizations that are motivated by these goals: a cross-gradient and a normalized cross-gradient structural similarity term, the vectorial total variation, and a joint regularization based on the nuclear norm of the gradients. Based on numerical results from three classes of inverse problems with piecewise-homogeneous parameter fields, we conclude that the vectorial total variation functional is preferable to the other methods considered. Besides resulting in good reconstructions in all experiments, it allows for scalable, efficient solvers for joint inverse problems governed by PDE forward models.

AB - Joint inversion refers to the simultaneous inference of multiple parameter fields from observations of systems governed by single or multiple forward models. In many cases these parameter fields reflect different attributes of a single medium and are thus spatially correlated or structurally similar. By imposing prior information on their spatial correlations via a joint regularization term, we seek to improve the reconstruction of the parameter fields relative to inversion for each field independently. One of the main challenges is to devise a joint regularization functional that conveys the spatial correlations or structural similarity between the fields while at the same time permitting scalable and efficient solvers for the joint inverse problem. We describe several joint regularizations that are motivated by these goals: a cross-gradient and a normalized cross-gradient structural similarity term, the vectorial total variation, and a joint regularization based on the nuclear norm of the gradients. Based on numerical results from three classes of inverse problems with piecewise-homogeneous parameter fields, we conclude that the vectorial total variation functional is preferable to the other methods considered. Besides resulting in good reconstructions in all experiments, it allows for scalable, efficient solvers for joint inverse problems governed by PDE forward models.

KW - cross-gradient

KW - joint inversion

KW - multi-physics inverse problem

KW - nuclear norm

KW - structural similarity prior

KW - vectorial total variation

UR - http://www.scopus.com/inward/record.url?scp=85062562660&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062562660&partnerID=8YFLogxK

U2 - 10.1088/1361-6420/aaf129

DO - 10.1088/1361-6420/aaf129

M3 - Article

VL - 35

JO - Inverse Problems

JF - Inverse Problems

SN - 0266-5611

IS - 2

M1 - 024003

ER -