Deep learning reconstruction for single pixel imaging with generative adversarial networks
buir.contributor.author | Güven, Baturalp | |
buir.contributor.author | Çukur, Tolga | |
buir.contributor.orcid | Çukur, Tolga|0000-0002-2296-851X | |
dc.citation.epage | 2064 | en_US |
dc.citation.spage | 2060 | |
dc.contributor.author | Güven, Baturalp | |
dc.contributor.author | Güngör, A. | |
dc.contributor.author | Bahçeci, M. U. | |
dc.contributor.author | Çukur, Tolga | |
dc.coverage.spatial | Kuala Lumpur, Malaysia | |
dc.date.accessioned | 2024-03-07T10:38:14Z | |
dc.date.available | 2024-03-07T10:38:14Z | |
dc.date.issued | 2023-09-11 | |
dc.department | Department of Electrical and Electronics Engineering | |
dc.description | Conference Name: 2023 IEEE International Conference on Image Processing (ICIP) | |
dc.description | Date of Conference: 08-11 October 2023 | |
dc.description.abstract | Single pixel imaging (SPI) enables high-resolution imaging through multiple coded measurements based on low-resolution snapshots. An inverse problem can then be solved to reconstruct a high-resolution image given the coded measurements. There has been recent interest in adoption of deep neural networks in SPI reconstruction. However, existing methods are commonly trained with pixel-wise loss terms such as the ℓ 1 -norm loss, which can result in spatial blurring and poor sensitivity to structural details. In this study, we propose a novel approach for deep SPI reconstruction based on an unrolled conditional generative adversarial network (cGAN) model. The generator estimates the high-resolution image using coded low-resolution measurements by iterating across a cascade of denoising and data-consistency modules. Meanwhile, the discriminator distinguishes real versus synthesized high-resolution images. The architecture is trained end-to-end via a combined pixel-wise and adversarial loss to enhance sensitivity to structural details. The proposed method is demonstrated against existing SPI reconstruction methods, and ablation studies are performed to demonstrate the individual model components. The proposed method outperforms competing methods in terms of both quantitative metrics and visual quality. | |
dc.description.provenance | Made available in DSpace on 2024-03-07T10:38:14Z (GMT). No. of bitstreams: 1 Deep_learning_reconstruction_for_single_pixel_imaging_with_generative_adversarial_networks.pdf: 4163822 bytes, checksum: e941effa80a0c4b8192166137c695d4d (MD5) Previous issue date: 2023-09-11 | en |
dc.identifier.doi | 10.1109/ICIP49359.2023.10223149 | |
dc.identifier.eisbn | 978-1-7281-9835-4 | |
dc.identifier.isbn | 978-1-7281-9836-1 | |
dc.identifier.uri | https://hdl.handle.net/11693/114387 | |
dc.language.iso | en_US | |
dc.publisher | IEEE | |
dc.relation.isversionof | https://dx.doi.org/10.1109/ICIP49359.2023.10223149 | |
dc.source.title | 2023 IEEE International Conference on Image Processing (ICIP) | |
dc.subject | Conditional generative adversarial network (cGAN) | |
dc.subject | Single pixel imaging (SPI) | |
dc.subject | Adversarial loss | |
dc.subject | Plug and play | |
dc.subject | ℓ1 Loss | |
dc.title | Deep learning reconstruction for single pixel imaging with generative adversarial networks | |
dc.type | Conference Paper |
Files
Original bundle
1 - 1 of 1
Loading...
- Name:
- Deep_learning_reconstruction_for_single_pixel_imaging_with_generative_adversarial_networks.pdf
- Size:
- 3.97 MB
- Format:
- Adobe Portable Document Format
License bundle
1 - 1 of 1
No Thumbnail Available
- Name:
- license.txt
- Size:
- 2.01 KB
- Format:
- Item-specific license agreed upon to submission
- Description: