Feature-Domain Proximal High-Dimensional Gradient Descent Network for Image Compressed Sensing

Recently, by incorporating optimization theory into deep neural networks, optimization-inspired networks have achieved remarkable success in image compressed sensing (ICS). However, existing networks only calculate a single gradient at each phase, making it difficult to fully utilize the measurement...

Full description

Saved in:
Bibliographic Details
Published in2023 IEEE International Conference on Image Processing (ICIP) pp. 1475 - 1479
Main Authors Liang, Ziwen, Yang, Chunling
Format Conference Proceeding
LanguageEnglish
Published IEEE 08.10.2023
Subjects
Artificial neural networks
Compressed sensing
deep learning
Image coding
image compressed sensing
Image restoration
Integrated circuits
inverse problem
Phase measurement
Sensors
Online AccessGet full text

Cover

More Information
Summary:Recently, by incorporating optimization theory into deep neural networks, optimization-inspired networks have achieved remarkable success in image compressed sensing (ICS). However, existing networks only calculate a single gradient at each phase, making it difficult to fully utilize the measurement information. In this paper, we propose a novel idea of calculating gradient in high-dimensional space during the updating process to fully exploit measurement information. On this basis, we further propose the Feature-domain Proximal High-dimensional Gradient Descent (FPHGD) algorithm to realize the proximal gradient descent in feature-domain and design a Feature-domain Proximal High-dimensional Gradient Descent network (FPHGD-Net) for ICS. Besides, to meet different requirements, we give three designs that combine different proximal mapping patterns in feature-domain to construct the networks. Extensive experiments demonstrate that reconstruction performance of our networks outperform existing state-of-the-art methods.
DOI:10.1109/ICIP49359.2023.10222347