SALSA-Net: Explainable Deep Unrolling Networks for Compressed Sensing
Deep unrolling networks (DUNs) have emerged as a promising approach for solving compressed sensing (CS) problems due to their superior explainability, speed, and performance compared to classical deep network models. However, the CS performance in terms of efficiency and accuracy remains a principal...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-05-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/23/11/5142 |
| _version_ | 1797596795066384384 |
|---|---|
| author | Heping Song Qifeng Ding Jingyao Gong Hongying Meng Yuping Lai |
| author_facet | Heping Song Qifeng Ding Jingyao Gong Hongying Meng Yuping Lai |
| author_sort | Heping Song |
| collection | DOAJ |
| description | Deep unrolling networks (DUNs) have emerged as a promising approach for solving compressed sensing (CS) problems due to their superior explainability, speed, and performance compared to classical deep network models. However, the CS performance in terms of efficiency and accuracy remains a principal challenge for approaching further improvements. In this paper, we propose a novel deep unrolling model, SALSA-Net, to solve the image CS problem. The network architecture of SALSA-Net is inspired by unrolling and truncating the split augmented Lagrangian shrinkage algorithm (SALSA) which is used to solve sparsity-induced CS reconstruction problems. SALSA-Net inherits the interpretability of the SALSA algorithm while incorporating the learning ability and fast reconstruction speed of deep neural networks. By converting the SALSA algorithm into a deep network structure, SALSA-Net consists of a gradient update module, a threshold denoising module, and an auxiliary update module. All parameters, including the shrinkage thresholds and gradient steps, are optimized through end-to-end learning and are subject to forward constraints to ensure faster convergence. Furthermore, we introduce learned sampling to replace traditional sampling methods so that the sampling matrix can better preserve the feature information of the original signal and improve sampling efficiency. Experimental results demonstrate that SALSA-Net achieves significant reconstruction performance compared to state-of-the-art methods while inheriting the advantages of explainable recovery and high speed from the DUNs paradigm. |
| first_indexed | 2024-03-11T02:58:04Z |
| format | Article |
| id | doaj.art-d02498e01f6549b3938cd2b3e4754f08 |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-03-11T02:58:04Z |
| publishDate | 2023-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-d02498e01f6549b3938cd2b3e4754f082023-11-18T08:33:07ZengMDPI AGSensors1424-82202023-05-012311514210.3390/s23115142SALSA-Net: Explainable Deep Unrolling Networks for Compressed SensingHeping Song0Qifeng Ding1Jingyao Gong2Hongying Meng3Yuping Lai4School of Computer Science and Communication Engineering, Jiangsu University, Zhenjiang 212013, ChinaSchool of Computer Science and Communication Engineering, Jiangsu University, Zhenjiang 212013, ChinaSchool of Computer Science and Communication Engineering, Jiangsu University, Zhenjiang 212013, ChinaElectronic and Electrical Engineering Department, Brunel University London, Uxbridge UB8 3PH, UKSchool of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing 100876, ChinaDeep unrolling networks (DUNs) have emerged as a promising approach for solving compressed sensing (CS) problems due to their superior explainability, speed, and performance compared to classical deep network models. However, the CS performance in terms of efficiency and accuracy remains a principal challenge for approaching further improvements. In this paper, we propose a novel deep unrolling model, SALSA-Net, to solve the image CS problem. The network architecture of SALSA-Net is inspired by unrolling and truncating the split augmented Lagrangian shrinkage algorithm (SALSA) which is used to solve sparsity-induced CS reconstruction problems. SALSA-Net inherits the interpretability of the SALSA algorithm while incorporating the learning ability and fast reconstruction speed of deep neural networks. By converting the SALSA algorithm into a deep network structure, SALSA-Net consists of a gradient update module, a threshold denoising module, and an auxiliary update module. All parameters, including the shrinkage thresholds and gradient steps, are optimized through end-to-end learning and are subject to forward constraints to ensure faster convergence. Furthermore, we introduce learned sampling to replace traditional sampling methods so that the sampling matrix can better preserve the feature information of the original signal and improve sampling efficiency. Experimental results demonstrate that SALSA-Net achieves significant reconstruction performance compared to state-of-the-art methods while inheriting the advantages of explainable recovery and high speed from the DUNs paradigm.https://www.mdpi.com/1424-8220/23/11/5142compressed sensingSALSAdeep unrollingexplainable networksneural networksimage reconstruction |
| spellingShingle | Heping Song Qifeng Ding Jingyao Gong Hongying Meng Yuping Lai SALSA-Net: Explainable Deep Unrolling Networks for Compressed Sensing Sensors compressed sensing SALSA deep unrolling explainable networks neural networks image reconstruction |
| title | SALSA-Net: Explainable Deep Unrolling Networks for Compressed Sensing |
| title_full | SALSA-Net: Explainable Deep Unrolling Networks for Compressed Sensing |
| title_fullStr | SALSA-Net: Explainable Deep Unrolling Networks for Compressed Sensing |
| title_full_unstemmed | SALSA-Net: Explainable Deep Unrolling Networks for Compressed Sensing |
| title_short | SALSA-Net: Explainable Deep Unrolling Networks for Compressed Sensing |
| title_sort | salsa net explainable deep unrolling networks for compressed sensing |
| topic | compressed sensing SALSA deep unrolling explainable networks neural networks image reconstruction |
| url | https://www.mdpi.com/1424-8220/23/11/5142 |
| work_keys_str_mv | AT hepingsong salsanetexplainabledeepunrollingnetworksforcompressedsensing AT qifengding salsanetexplainabledeepunrollingnetworksforcompressedsensing AT jingyaogong salsanetexplainabledeepunrollingnetworksforcompressedsensing AT hongyingmeng salsanetexplainabledeepunrollingnetworksforcompressedsensing AT yupinglai salsanetexplainabledeepunrollingnetworksforcompressedsensing |