Deep Mr Image Super-Resolution Using Structural Priors

Venkateswararao Cherukuri, Tiantong Guo, Steven Schiff, Vishal Monga

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

High resolution magnetic resonance (MR) images are desired for accurate diagnostics. In practice, image resolution is restricted by factors like hardware, cost and processing constraints. Recently, deep learning methods have been shown to produce compelling state of the art results for image super-resolution. Paying particular attention to desired hi-resolution MR image structure, we propose a new regularized network that exploits image priors, namely a low-rank structure and a sharpness prior to enhance deep MR image superresolution. Our contributions are then incorporating these priors in an analytically tractable fashion in the learning of a convolutional neural network (CNN) that accomplishes the super-resolution task. This is particularly challenging for the low rank prior, since the rank is not a differentiable function of the image matrix (and hence the network parameters), an issue we address by pursuing differentiable approximations of the rank. Sharpness is emphasized by the variance of the Laplacian which we show can be implemented by a fixedJeedback layer at the output of the network. Experiments performed on two publicly available MR brain image databases exhibit promising results particularly when training imagery is limited.

Original languageEnglish (US)
Title of host publication2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings
PublisherIEEE Computer Society
Pages410-414
Number of pages5
ISBN (Electronic)9781479970612
DOIs
StatePublished - Aug 29 2018
Event25th IEEE International Conference on Image Processing, ICIP 2018 - Athens, Greece
Duration: Oct 7 2018Oct 10 2018

Publication series

NameProceedings - International Conference on Image Processing, ICIP
ISSN (Print)1522-4880

Conference

Conference25th IEEE International Conference on Image Processing, ICIP 2018
CountryGreece
CityAthens
Period10/7/1810/10/18

Fingerprint

Magnetic resonance
Image resolution
Brain
Neural networks
Hardware
Processing
Costs
Experiments

All Science Journal Classification (ASJC) codes

  • Software
  • Computer Vision and Pattern Recognition
  • Signal Processing

Cite this

Cherukuri, V., Guo, T., Schiff, S., & Monga, V. (2018). Deep Mr Image Super-Resolution Using Structural Priors. In 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings (pp. 410-414). [8451496] (Proceedings - International Conference on Image Processing, ICIP). IEEE Computer Society. https://doi.org/10.1109/ICIP.2018.8451496
Cherukuri, Venkateswararao ; Guo, Tiantong ; Schiff, Steven ; Monga, Vishal. / Deep Mr Image Super-Resolution Using Structural Priors. 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings. IEEE Computer Society, 2018. pp. 410-414 (Proceedings - International Conference on Image Processing, ICIP).
@inproceedings{02a8c97e0e76486d9f171195bf9d36bd,
title = "Deep Mr Image Super-Resolution Using Structural Priors",
abstract = "High resolution magnetic resonance (MR) images are desired for accurate diagnostics. In practice, image resolution is restricted by factors like hardware, cost and processing constraints. Recently, deep learning methods have been shown to produce compelling state of the art results for image super-resolution. Paying particular attention to desired hi-resolution MR image structure, we propose a new regularized network that exploits image priors, namely a low-rank structure and a sharpness prior to enhance deep MR image superresolution. Our contributions are then incorporating these priors in an analytically tractable fashion in the learning of a convolutional neural network (CNN) that accomplishes the super-resolution task. This is particularly challenging for the low rank prior, since the rank is not a differentiable function of the image matrix (and hence the network parameters), an issue we address by pursuing differentiable approximations of the rank. Sharpness is emphasized by the variance of the Laplacian which we show can be implemented by a fixedJeedback layer at the output of the network. Experiments performed on two publicly available MR brain image databases exhibit promising results particularly when training imagery is limited.",
author = "Venkateswararao Cherukuri and Tiantong Guo and Steven Schiff and Vishal Monga",
year = "2018",
month = "8",
day = "29",
doi = "10.1109/ICIP.2018.8451496",
language = "English (US)",
series = "Proceedings - International Conference on Image Processing, ICIP",
publisher = "IEEE Computer Society",
pages = "410--414",
booktitle = "2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings",
address = "United States",

}

Cherukuri, V, Guo, T, Schiff, S & Monga, V 2018, Deep Mr Image Super-Resolution Using Structural Priors. in 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings., 8451496, Proceedings - International Conference on Image Processing, ICIP, IEEE Computer Society, pp. 410-414, 25th IEEE International Conference on Image Processing, ICIP 2018, Athens, Greece, 10/7/18. https://doi.org/10.1109/ICIP.2018.8451496

Deep Mr Image Super-Resolution Using Structural Priors. / Cherukuri, Venkateswararao; Guo, Tiantong; Schiff, Steven; Monga, Vishal.

2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings. IEEE Computer Society, 2018. p. 410-414 8451496 (Proceedings - International Conference on Image Processing, ICIP).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Deep Mr Image Super-Resolution Using Structural Priors

AU - Cherukuri, Venkateswararao

AU - Guo, Tiantong

AU - Schiff, Steven

AU - Monga, Vishal

PY - 2018/8/29

Y1 - 2018/8/29

N2 - High resolution magnetic resonance (MR) images are desired for accurate diagnostics. In practice, image resolution is restricted by factors like hardware, cost and processing constraints. Recently, deep learning methods have been shown to produce compelling state of the art results for image super-resolution. Paying particular attention to desired hi-resolution MR image structure, we propose a new regularized network that exploits image priors, namely a low-rank structure and a sharpness prior to enhance deep MR image superresolution. Our contributions are then incorporating these priors in an analytically tractable fashion in the learning of a convolutional neural network (CNN) that accomplishes the super-resolution task. This is particularly challenging for the low rank prior, since the rank is not a differentiable function of the image matrix (and hence the network parameters), an issue we address by pursuing differentiable approximations of the rank. Sharpness is emphasized by the variance of the Laplacian which we show can be implemented by a fixedJeedback layer at the output of the network. Experiments performed on two publicly available MR brain image databases exhibit promising results particularly when training imagery is limited.

AB - High resolution magnetic resonance (MR) images are desired for accurate diagnostics. In practice, image resolution is restricted by factors like hardware, cost and processing constraints. Recently, deep learning methods have been shown to produce compelling state of the art results for image super-resolution. Paying particular attention to desired hi-resolution MR image structure, we propose a new regularized network that exploits image priors, namely a low-rank structure and a sharpness prior to enhance deep MR image superresolution. Our contributions are then incorporating these priors in an analytically tractable fashion in the learning of a convolutional neural network (CNN) that accomplishes the super-resolution task. This is particularly challenging for the low rank prior, since the rank is not a differentiable function of the image matrix (and hence the network parameters), an issue we address by pursuing differentiable approximations of the rank. Sharpness is emphasized by the variance of the Laplacian which we show can be implemented by a fixedJeedback layer at the output of the network. Experiments performed on two publicly available MR brain image databases exhibit promising results particularly when training imagery is limited.

UR - http://www.scopus.com/inward/record.url?scp=85062897175&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062897175&partnerID=8YFLogxK

U2 - 10.1109/ICIP.2018.8451496

DO - 10.1109/ICIP.2018.8451496

M3 - Conference contribution

C2 - 30930696

AN - SCOPUS:85062897175

T3 - Proceedings - International Conference on Image Processing, ICIP

SP - 410

EP - 414

BT - 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings

PB - IEEE Computer Society

ER -

Cherukuri V, Guo T, Schiff S, Monga V. Deep Mr Image Super-Resolution Using Structural Priors. In 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings. IEEE Computer Society. 2018. p. 410-414. 8451496. (Proceedings - International Conference on Image Processing, ICIP). https://doi.org/10.1109/ICIP.2018.8451496