Learning Regularization Parameter-Maps for Variational Image Reconstruction Using Deep Neural Networks and Algorithm Unrolling
From MaRDI portal
Publication:6144067
Abstract: We introduce a method for fast estimation of data-adapted, spatio-temporally dependent regularization parameter-maps for variational image reconstruction, focusing on total variation (TV)-minimization. Our approach is inspired by recent developments in algorithm unrolling using deep neural networks (NNs), and relies on two distinct sub-networks. The first sub-network estimates the regularization parameter-map from the input data. The second sub-network unrolls iterations of an iterative algorithm which approximately solves the corresponding TV-minimization problem incorporating the previously estimated regularization parameter-map. The overall network is trained end-to-end in a supervised learning fashion using pairs of clean-corrupted data but crucially without the need of having access to labels for the optimal regularization parameter-maps. We prove consistency of the unrolled scheme by showing that the unrolled energy functional used for the supervised learning -converges as tends to infinity, to the corresponding functional that incorporates the exact solution map of the TV-minimization problem. We apply and evaluate our method on a variety of large scale and dynamic imaging problems in which the automatic computation of such parameters has been so far challenging: 2D dynamic cardiac MRI reconstruction, quantitative brain MRI reconstruction, low-dose CT and dynamic image denoising. The proposed method consistently improves the TV-reconstructions using scalar parameters and the obtained parameter-maps adapt well to each imaging problem and data by leading to the preservation of detailed features. Although the choice of the regularization parameter-maps is data-driven and based on NNs, the proposed algorithm is entirely interpretable since it inherits the properties of the respective iterative reconstruction method from which the network is implicitly defined.
Recommendations
- Iteratively Refined Image Reconstruction with Learned Attentive Regularizers
- Variational model-based deep neural networks for image reconstruction
- Learned regularizers for inverse problems
- Deep unfolding of a proximal interior point method for image restoration
- Learning maximally monotone operators for image recovery
Cites work
- A bilevel optimization approach for parameter learning in variational models
- A Fast Adaptive Parameter Estimation for Total Variation Image Restoration
- A first-order primal-dual algorithm for convex problems with applications to imaging
- A guide to the TV zoo
- A New Alternating Minimization Algorithm for Total Variation Image Reconstruction
- A new primal-dual algorithm for minimizing the sum of three functions with a linear operator
- Adaptive wavelet thresholding for image denoising and compression
- An Infeasible Primal-Dual Algorithm for Total Bounded Variation--Based Inf-Convolution-Type Image Restoration
- An introduction to -convergence
- An introduction to continuous optimization for imaging
- An Unbiased Risk Estimator for Image Denoising in the Presence of Mixed Poisson–Gaussian Noise
- Analytical aspects of spatially adapted total variation regularisation
- Augmented Lagrangian method, dual methods, and split Bregman iteration for ROF, vectorial TV, and high order models
- Bilevel approaches for learning of variational imaging models
- Bilevel Methods for Image Reconstruction
- Bilevel optimization with nonsmooth lower level problems
- Bilevel parameter learning for higher-order total variation regularisation models
- Bilevel parameter learning for nonlocal image denoising models
- Bilevel Training Schemes in Imaging for Total Variation--Type Functionals with Convex Integrands
- Dualization and Automatic Distributed Parameter Selection of Total Generalized Variation via Bilevel Optimization
- Efficient numerical methods in non-uniform sampling theory
- Generalized Cross-Validation as a Method for Choosing a Good Ridge Parameter
- Generating structured nonsmooth priors and associated primal-dual methods
- scientific article; zbMATH DE number 3227378 (Why is no real title available?)
- Ideal spatial adaptation by wavelet shrinkage
- Image denoising: learning the noise model via nonsmooth PDE-constrained optimization
- Image recovery via total variation minimization and related problems
- Learning optimal spatially-dependent regularization parameters in total variation image denoising
- Learning regularization parameters of inverse problems via deep neural networks
- Model selection with low complexity priors
- NETT: solving inverse problems with deep neural networks
- Nonlinear total variation based noise removal algorithms
- On infimal convolution of TV-type functionals and applications to video and image reconstruction
- One condition for solution uniqueness and robustness of both _1-synthesis and _1-analysis minimizations
- Optimal selection of the regularization function in a weighted total variation model. I: Modelling and theory
- Optimal selection of the regularization function in a weighted total variation model. II: Algorithm, its analysis and numerical tests
- Optimality Conditions for Bilevel Imaging Learning Problems with Total Variation Regularization
- Robust Sparse Analysis Regularization
- Solving ill-posed inverse problems using iterative deep neural networks
- Solving inverse problems using data-driven models
- The Discontinuity Set of Solutions of the TV Denoising Problem and Some Extensions
- The Little Engine that Could: Regularization by Denoising (RED)
- The Split Bregman Method for L1-Regularized Problems
- Tikhonov regularization and the L-curve for large discrete ill-posed problems
- Total Bounded Variation Regularization as a Bilaterally Constrained Optimization Problem
- Total generalized variation
- Variational methods in imaging
Cited in
(17)- Regularization by architecture: a deep prior approach for inverse problems
- Reversible Autoencoder: A CNN-Based Nonlinear Lifting Scheme for Image Reconstruction
- Self-Supervised Deep Learning for Image Reconstruction: A Langevin Monte Carlo Approach
- Speckle noise removal via learned variational models
- Variational model-based deep neural networks for image reconstruction
- Iteratively Refined Image Reconstruction with Learned Attentive Regularizers
- Adaptive weighted total variation boosted by learning techniques in few-view tomographic imaging
- An adaptively inexact method for bilevel learning using primal-dual-style differentiation
- Deep unrolling for learning optimal spatially varying regularisation parameters for total generalised variation
- Whiteness-based bilevel estimation of weighted TV parameter maps for image denoising
- Theoretical characterization of effect of masks in snapshot compressive imaging
- Stability of data-dependent ridge-regularization for inverse problems
- Data-driven methods for quantitative imaging
- Dyadic partition-based training schemes for TV/TGV denoising
- Algorithmic unfolding for image reconstruction and localization problems in single-molecule fluorescence microscopy
- Neural networks for Bayesian inverse problems governed by a nonlinear ODE
- Adversarially learned iterative reconstruction for imaging inverse problems
This page was built for publication: Learning Regularization Parameter-Maps for Variational Image Reconstruction Using Deep Neural Networks and Algorithm Unrolling
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6144067)
