Convolutional Sparse Coding Network Via Improved Proximal Gradient For Compressed Sensing Magnetic Resonance Imaging
Xiaofan Wang (Yanshan University), Yali Zhang (Yanshan University), Pengyu Li (Yanshan University), Jinjia Wang (Yanshan University)*
The 33rd British Machine Vision Conference
Abstract
Compressed sensing magnetic resonance imaging (CS-MRI) applies compressed sensing (CS) to effectively accelerate image reconstruction from undersampling k-space data. Compared with the traditional patch-based sparse representation of CS, the slice-based convolutional sparse coding (CSC) model makes up for the shortcomings of the sparse coding by establishing translation invariance. In this paper, based on the slice-based CSC model, we propose an improved proximal gradient algorithm to optimize image reconstruction performance. First, the variance regularization term is introduced into the CSC problem to remove the constraint on the convolutional dictionary. Second, we introduce the heavy ball system with dry friction from the dynamic system perspective to find a better local optimal solution. Then, the improved proximal gradient algorithm is unfolded into an encoder network to obtain the coding. The reconstruction process is modeled as a decoder network. The convolutional dictionary is updated by the backpropagation algorithm via the mean square error of the reconstructed signal. Compared with the current methods and the latest network, the proposed model-based novel network is demonstrated that it achieves better reconstruction performances.Video
Citation
@inproceedings{Wang_2022_BMVC,
author = {Xiaofan Wang and Yali Zhang and Pengyu Li and Jinjia Wang},
title = {Convolutional Sparse Coding Network Via Improved Proximal Gradient For Compressed Sensing Magnetic Resonance Imaging},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year = {2022},
url = {https://bmvc2022.mpi-inf.mpg.de/0090.pdf}
}
Copyright © 2022
The British Machine Vision Conference is organised by