Convolutional Neural Networks for Inverse Problems in Imaging: A Review

By Michael T. McCann, Kyong Hwan Jin, Michael A. Unser. 2017

665 citations recorded by the source corpus.

In this article, we review recent uses of convolutional neural networks (CNNs) to solve inverse problems in imaging. It has recently become feasible to train deep CNNs on large databases of images, and they have shown outstanding performance on object classification and segmentation tasks. Motivated by these successes, researchers have begun to apply CNNs to the resolution of inverse problems such as denoising, deconvolution, superresolution, and medical image reconstruction, and they have started to report improvements over state-of-the-art methods, including sparsity-based techniques such as compressed sensing. Here, we review the recent experimental work in these areas, with a focus on the critical design decisions.

Subfields: Artificial Intelligence, Computational Mechanics, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Gregory C. Beroza
2. Yunjie Yang
3. P. Johnson
4. Richard Baraniuk
5. Andreas K. Maier
6. Christian Riess
7. Arvid Lundervold
8. Rebecca M. Willett

In Computational Mechanics

1. A. Dimakis
2. Christopher A. Metzler
3. Richard Baraniuk
4. Ajil Jalal
5. Rebecca M. Willett
6. Arvid Lundervold
7. Greg Ongie
8. Yunjie Yang

In Computer Vision and Pattern Recognition

1. A. Dimakis
2. Ajil Jalal
3. Richard Baraniuk
4. Arvid Lundervold
5. Christopher A. Metzler
6. Michael T. McCann
7. Kyong Hwan Jin
8. Michael A. Unser

Citations in the corpus

• An overview of deep learning in medical imaging focusing on MRI (2018, background): “(30)See [154, 155, 156, 157, 158] for recent perspectives and developments connecting deep learning-based reconstruction methods to the more general research field of inverse problems.”
• Deep Learning on Image Denoising: An overview (2019, neutral): “For this reason, CNNs have been developed for image denoising [155, 134].”
• FISTA-Net: Learning a Fast Iterative Shrinkage Thresholding Network for Inverse Problems in Imaging (2020, background): “In this approach, the deep learning model must learn the underlying physics of the problem, which is difficult or even infeasible [18].”
• Machine learning for data-driven discovery in solid Earth geoscience (2019, background): “ML, especially deep learning, can aid in the analysis of inverse problems ( 26 ).”
• Deep Learning Techniques for Inverse Problems in Imaging (2020, background): “For example, in super-resolution a common choice of Ã−1 is upsampling by bicubic interpolation [18]; in CT reconstruction, a common choice of Ã−1 is filtered back projection [98].”
• A Gentle Introduction to Deep Learning in Medical Image Processing (2018, background): “Recent developments in this image-to-image reconstruction are summarized in [103].”
• CNN-Based Projected Gradient Descent for Consistent CT Image Reconstruction (2017, background): “3) Learning-Based Algorithms: Recently, a surge in using deep learning to solve inverse problems in imaging [15]–[19], has established new state-of-the-art results for tasks such as sparse-view CT reconstruction [16].”