Deep Unfolding of a Proximal Interior Point Method for Image Restoration - Archive ouverte HAL Access content directly
Journal Articles Inverse Problems Year : 2020

Deep Unfolding of a Proximal Interior Point Method for Image Restoration

Abstract

Variational methods are widely applied to ill-posed inverse problems for they have the ability to embed prior knowledge about the solution. However, the level of performance of these methods significantly depends on a set of parameters, which can be estimated through computationally expensive and time-consuming methods. In contrast, deep learning offers very generic and efficient architectures, at the expense of explainability, since it is often used as a black-box, without any fine control over its output. Deep unfolding provides a convenient approach to combine variational-based and deep learning approaches. Starting from a variational formulation for image restoration, we develop iRestNet, a neural network architecture obtained by unfolding a proximal interior point algorithm. Hard constraints, encoding desirable properties for the restored image, are incorporated into the network thanks to a logarithmic barrier, while the barrier parameter, the stepsize, and the penalization weight are learned by the network. We derive explicit expressions for the gradient of the proximity operator for various choices of constraints, which allows training iRestNet with gradient descent and backpropagation. In addition, we provide theoretical results regarding the stability of the network for a common inverse problem example. Numerical experiments on image deblurring problems show that the proposed approach compares favorably with both state-of-the-art variational and machine learning methods in terms of image quality.
Fichier principal
Vignette du fichier
DeepUnfolding.pdf (2.76 Mo) Télécharger le fichier
1_gamma.pdf (16.4 Ko) Télécharger le fichier
1_lambda.pdf (18.35 Ko) Télécharger le fichier
1_mu.pdf (20.88 Ko) Télécharger le fichier
35028_gamma.pdf (15.95 Ko) Télécharger le fichier
35028_lambda.pdf (18.43 Ko) Télécharger le fichier
35028_mu.pdf (20.15 Ko) Télécharger le fichier
87015_gamma.pdf (18.96 Ko) Télécharger le fichier
87015_lambda.pdf (19.16 Ko) Télécharger le fichier
87015_mu.pdf (18.24 Ko) Télécharger le fichier
_ircnn_1_906_fcnn_1_856_irestnet_1_909.pdf (215.36 Ko) Télécharger le fichier
_ircnn_87015_685_irestnet_87015_713.pdf (149.57 Ko) Télécharger le fichier
_mlp_35028_860_pdhg_35028_772_ircnn_35028_840_irestnet_35028_883.pdf (243.99 Ko) Télécharger le fichier
cnnpp_ircnn.pdf (81.36 Ko) Télécharger le fichier
gaussian_1_6_std_0008_BSD500_res.pdf (46.97 Ko) Télécharger le fichier
gaussian_1_6_std_001_005_BSD500_res.pdf (47.06 Ko) Télécharger le fichier
gaussian_3_std_004_BSD500_res.pdf (47.3 Ko) Télécharger le fichier
groundtruth_1_blurred_1_576_var_1_844_epll_1_849.pdf (303.38 Ko) Télécharger le fichier
groundtruth_35028_blurred_35028_509_var_35028_833_epll_35028_839.pdf (351.26 Ko) Télécharger le fichier
groundtruth_87015_blurred_87015_344_var_87015_622_epll_87015_553.pdf (456.07 Ko) Télécharger le fichier
iopart10.clo (4.59 Ko) Télécharger le fichier
iopart12.clo (4.59 Ko) Télécharger le fichier
lkmu.pdf (82.11 Ko) Télécharger le fichier
motion3_std_001_BSD500_res.pdf (45.42 Ko) Télécharger le fichier
motion8_std_001_BSD500_res.pdf (50.72 Ko) Télécharger le fichier
net_archi.pdf (67.17 Ko) Télécharger le fichier
prox_barrier_l2norm_2D.pdf (61.61 Ko) Télécharger le fichier
prox_barrier_range.pdf (122.64 Ko) Télécharger le fichier
square_7_std_001_BSD500_res.pdf (41.65 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-01943475 , version 1 (07-12-2018)
hal-01943475 , version 2 (15-07-2019)
hal-01943475 , version 3 (16-07-2019)
hal-01943475 , version 4 (21-01-2020)

Identifiers

Cite

Carla Bertocchi, Emilie Chouzenoux, Marie-Caroline Corbineau, Jean-Christophe Pesquet, Marco Prato. Deep Unfolding of a Proximal Interior Point Method for Image Restoration. Inverse Problems, In press, ⟨10.1088/1361-6420/ab460a⟩. ⟨hal-01943475v4⟩
583 View
1193 Download

Altmetric

Share

Gmail Facebook X LinkedIn More