Deep Post-Processing for Sparse Image Deconvolution

Matthieu Terris, Abdullah Abdulaziz, Arwa Dabbech, Ming Jiang, Audrey Repetti, Jean-Christophe Pesquet, Yves Wiaux

Research output: Contribution to conferenceAbstract

Abstract

Variational-based methods are the state-of-the-art in sparse
image deconvolution. Yet, this class of methods might not scale to large
dimensions of interest in current high resolution imaging applications.
To overcome this limitation, we propose to solve the sparse deconvolution
problem through a two-step approach consisting in first solving
(approximately and fast) an optimization problem following by a neural
network for ”Deep Post Processing” (DPP). We illustrate our method in
radio astronomy, where algorithms scalability is paramount due to the
extreme data dimensions. First results suggest that DPP is able to achieve
similar quality to state-of-the-art methods in a fraction of the time.
Original languageEnglish
Number of pages2
Publication statusPublished - 1 Jul 2019
EventSignal Processing with Adaptive Sparse Structured Representations (SPARS) workshop - Toulouse, France
Duration: 1 Jul 20194 Jul 2019

Workshop

WorkshopSignal Processing with Adaptive Sparse Structured Representations (SPARS) workshop
Abbreviated titleSPARS 2019
CountryFrance
CityToulouse
Period1/07/194/07/19

Fingerprint

Deconvolution
Astronomy
Processing
Scalability
Imaging techniques

Cite this

Terris, M., Abdulaziz, A., Dabbech, A., Jiang, M., Repetti, A., Pesquet, J-C., & Wiaux, Y. (2019). Deep Post-Processing for Sparse Image Deconvolution. Abstract from Signal Processing with Adaptive Sparse Structured Representations (SPARS) workshop, Toulouse, France.
Terris, Matthieu ; Abdulaziz, Abdullah ; Dabbech, Arwa ; Jiang, Ming ; Repetti, Audrey ; Pesquet, Jean-Christophe ; Wiaux, Yves. / Deep Post-Processing for Sparse Image Deconvolution. Abstract from Signal Processing with Adaptive Sparse Structured Representations (SPARS) workshop, Toulouse, France.2 p.
@conference{9000a1397fdd4e57854bd7585b77cade,
title = "Deep Post-Processing for Sparse Image Deconvolution",
abstract = "Variational-based methods are the state-of-the-art in sparseimage deconvolution. Yet, this class of methods might not scale to largedimensions of interest in current high resolution imaging applications.To overcome this limitation, we propose to solve the sparse deconvolutionproblem through a two-step approach consisting in first solving(approximately and fast) an optimization problem following by a neuralnetwork for ”Deep Post Processing” (DPP). We illustrate our method inradio astronomy, where algorithms scalability is paramount due to theextreme data dimensions. First results suggest that DPP is able to achievesimilar quality to state-of-the-art methods in a fraction of the time.",
author = "Matthieu Terris and Abdullah Abdulaziz and Arwa Dabbech and Ming Jiang and Audrey Repetti and Jean-Christophe Pesquet and Yves Wiaux",
year = "2019",
month = "7",
day = "1",
language = "English",
note = "Signal Processing with Adaptive Sparse Structured Representations (SPARS) workshop, SPARS 2019 ; Conference date: 01-07-2019 Through 04-07-2019",

}

Terris, M, Abdulaziz, A, Dabbech, A, Jiang, M, Repetti, A, Pesquet, J-C & Wiaux, Y 2019, 'Deep Post-Processing for Sparse Image Deconvolution' Signal Processing with Adaptive Sparse Structured Representations (SPARS) workshop, Toulouse, France, 1/07/19 - 4/07/19, .

Deep Post-Processing for Sparse Image Deconvolution. / Terris, Matthieu; Abdulaziz, Abdullah; Dabbech, Arwa; Jiang, Ming; Repetti, Audrey; Pesquet, Jean-Christophe; Wiaux, Yves.

2019. Abstract from Signal Processing with Adaptive Sparse Structured Representations (SPARS) workshop, Toulouse, France.

Research output: Contribution to conferenceAbstract

TY - CONF

T1 - Deep Post-Processing for Sparse Image Deconvolution

AU - Terris, Matthieu

AU - Abdulaziz, Abdullah

AU - Dabbech, Arwa

AU - Jiang, Ming

AU - Repetti, Audrey

AU - Pesquet, Jean-Christophe

AU - Wiaux, Yves

PY - 2019/7/1

Y1 - 2019/7/1

N2 - Variational-based methods are the state-of-the-art in sparseimage deconvolution. Yet, this class of methods might not scale to largedimensions of interest in current high resolution imaging applications.To overcome this limitation, we propose to solve the sparse deconvolutionproblem through a two-step approach consisting in first solving(approximately and fast) an optimization problem following by a neuralnetwork for ”Deep Post Processing” (DPP). We illustrate our method inradio astronomy, where algorithms scalability is paramount due to theextreme data dimensions. First results suggest that DPP is able to achievesimilar quality to state-of-the-art methods in a fraction of the time.

AB - Variational-based methods are the state-of-the-art in sparseimage deconvolution. Yet, this class of methods might not scale to largedimensions of interest in current high resolution imaging applications.To overcome this limitation, we propose to solve the sparse deconvolutionproblem through a two-step approach consisting in first solving(approximately and fast) an optimization problem following by a neuralnetwork for ”Deep Post Processing” (DPP). We illustrate our method inradio astronomy, where algorithms scalability is paramount due to theextreme data dimensions. First results suggest that DPP is able to achievesimilar quality to state-of-the-art methods in a fraction of the time.

M3 - Abstract

ER -

Terris M, Abdulaziz A, Dabbech A, Jiang M, Repetti A, Pesquet J-C et al. Deep Post-Processing for Sparse Image Deconvolution. 2019. Abstract from Signal Processing with Adaptive Sparse Structured Representations (SPARS) workshop, Toulouse, France.