Multi-Scale Residual Reconstruction Neural Network With Non-Local Constraint

Wan, Li; Liu, Fang; Jiao, Licheng; Hu, Fei

doi:10.1109/access.2019.2918593

Cited by 8 publications

(7 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multi-scale structures were utilized in [31]. There are three branches of sub-networks with different convolutional kernel sizes to extract information of different scales.…”

Section: Neural Network As Image Projectionsmentioning

confidence: 99%

A Review of Deep Learning Methods for Compressed Sensing Image Reconstruction and Its Medical Applications

Xie

2022

Electronics

View full text Add to dashboard Cite

Compressed sensing (CS) and its medical applications are active areas of research. In this paper, we review recent works using deep learning method to solve CS problem for images or medical imaging reconstruction including computed tomography (CT), magnetic resonance imaging (MRI) and positron-emission tomography (PET). We propose a novel framework to unify traditional iterative algorithms and deep learning approaches. In short, we define two projection operators toward image prior and data consistency, respectively, and any reconstruction algorithm can be decomposed to the two parts. Though deep learning methods can be divided into several categories, they all satisfies the framework. We built the relationship between different reconstruction methods of deep learning, and connect them to traditional methods through the proposed framework. It also indicates that the key to solve CS problem and its medical applications is how to depict the image prior. Based on the framework, we analyze the current deep learning methods and point out some important directions of research in the future.

show abstract

“…Multi-scale structures were utilized in [31]. There are three branches of sub-networks with different convolutional kernel sizes to extract information of different scales.…”

Section: Neural Network As Image Projectionsmentioning

confidence: 99%

A Review of Deep Learning Methods for Compressed Sensing Image Reconstruction and Its Medical Applications

Xie

2022

Electronics

View full text Add to dashboard Cite

show abstract

“…However, the fully connected network (FCN) utilized in SDA leads to a huge number of learnable parameters. To relieve this problem, several Convolutional Neural Networks (CNNs) based reconstruction methods [15], [16], [18], [34] are proposed, which usually build a direct mapping from the blocked measurements to the corresponding image blocks. However, these deep network-based CS algorithms usually bring about serious block artifacts [6], [35] (especially at low sampling rates) because of their block-by-block reconstruction.…”

Section: A Image Cs Reconstructionmentioning

confidence: 99%

“…By exploring the non-local self-similarity prior, several deep network-based CS schemes begin to utilize the non-local priors. For example, Li et al [34] propose a residual network with nonlocal constraint for image CS reconstruction, which considers the non-local self-similarity image prior and adds a non-local operation into the proposed network. While this CS network reconstructs the target image in a block-by-block manner from the measurements acquired by using the Gaussian random sampling matrix.…”

Section: B Non-local Self-similarity Image Priormentioning

confidence: 99%

“…While this CS network reconstructs the target image in a block-by-block manner from the measurements acquired by using the Gaussian random sampling matrix. Besides, the non-local operator in [34] only perceives the self-similar information inside the current image block, which ignores the correlations among different image blocks. Compared to [34], the differences of the proposed framework are as follows: 1) Instead of blockby-block reconstruction manner, the proposed CS network reconstructs the target image from the measurements in the whole image space.…”

Section: B Non-local Self-similarity Image Priormentioning

confidence: 99%

“…Besides, the non-local operator in [34] only perceives the self-similar information inside the current image block, which ignores the correlations among different image blocks. Compared to [34], the differences of the proposed framework are as follows: 1) Instead of blockby-block reconstruction manner, the proposed CS network reconstructs the target image from the measurements in the whole image space. 2) The sampling matrix of the proposed CS model is optimized jointly with the reconstruction network.…”

Section: B Non-local Self-similarity Image Priormentioning

confidence: 99%

See 2 more Smart Citations

Image Compressed Sensing Using Non-local Neural Network

Cui,

Liu,

Jiang

et al. 2021

Preprint

View full text Add to dashboard Cite

Deep network-based image Compressed Sensing (CS) has attracted much attention in recent years. However, the existing deep network-based CS schemes either reconstruct the target image in a block-by-block manner that leads to serious block artifacts or train the deep network as a black box that brings about limited insights of image prior knowledge. In this paper, a novel image CS framework using non-local neural network (NL-CSNet) is proposed, which utilizes the nonlocal self-similarity priors with deep network to improve the reconstruction quality. In the proposed NL-CSNet, two nonlocal subnetworks are constructed for utilizing the non-local selfsimilarity priors in the measurement domain and the multi-scale feature domain respectively. Specifically, in the subnetwork of measurement domain, the long-distance dependencies between the measurements of different image blocks are established for better initial reconstruction. Analogically, in the subnetwork of multi-scale feature domain, the affinities between the dense feature representations are explored in the multi-scale space for deep reconstruction. Furthermore, a novel loss function is developed to enhance the coupling between the non-local representations, which also enables an end-to-end training of NL-CSNet. Extensive experiments manifest that NL-CSNet outperforms existing state-of-the-art CS methods, while maintaining fast computational speed. The source code of plugging nonlocal module into certain CS network can be found: https: //github.com/WenxueCui/NL-CSNet-Pytorch.

show abstract

NL-CS Net: Deep Learning with Non-local Prior for Image Compressive Sensing

Bian,

Qi,

et al. 2024

Circuits Syst Signal Process

View full text Add to dashboard Cite

Multi-Scale Residual Reconstruction Neural Network With Non-Local Constraint

Cited by 8 publications

References 21 publications

A Review of Deep Learning Methods for Compressed Sensing Image Reconstruction and Its Medical Applications

A Review of Deep Learning Methods for Compressed Sensing Image Reconstruction and Its Medical Applications

Image Compressed Sensing Using Non-local Neural Network

NL-CS Net: Deep Learning with Non-local Prior for Image Compressive Sensing

Contact Info

Product

Resources

About