Research map / 3D vision / Depth and stereo
Deep Spatial-angular Regularization for Compressive Light Field Reconstruction over Coded Apertures
Mantang Guo, Junhui Hou, Jing Jin, Jie Chen, Lap-Pui Chau
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use the proposed network instead of Inagaki et al. | Ours (Single) · Kalantari Lytro, 1 to 49 | EPI-SSIM: 0.901 -> 0.935 | Improved |
| use the proposed network instead of Inagaki et al. | Ours (Single) · Kalantari Lytro, 2 to 49 | EPI-SSIM: 0.968 -> 0.981 | Improved |
| use the proposed network instead of Inagaki et al. | Ours (Single) · Kalantari Lytro, 4 to 49 | EPI-SSIM: 0.973 -> 0.986 | Improved |
| use the proposed network instead of Inagaki et al. | Ours (Single) · Kalantari Lytro, 1 to 49 | running time (s): 218.94 -> 80.71 | Improved |
| use the proposed network instead of Kalantari et al. | Ours (Single) · Kalantari Lytro, 4 to 49 | running time (s): 168.86 -> 80.54 | Improved |
| use the proposed network instead of Yeung et al. | Ours (Single) · Kalantari Lytro, 4 to 49 | running time (s): 0.85 -> 80.54 | Hurt |
| use three-channel (Multiple) learned aperture instead of shared single-channel kernel | Ours · Kalantari Lytro, all tasks | PSNR | Improved |
| increase iterative stages from 1 to 3 and 6 (ablation) | Ours · synthetic LF, task 2 to 25 | PSNR/SSIM: stages 1, 3, 6 | Improved |
| increase number of SAS convolutional layers from 3 to 6 and 9 (ablation) | Ours · synthetic LF, task 2 to 25 | PSNR/SSIM: SAS 3, 6, 9 | Improved |
| raise Gaussian noise on measurements to sigma = 30 (ablation) | Ours · synthetic LF, task 2 to 25 | PSNR: sigma 3, 6, 30 | Hurt |
| add mild Gaussian noise (sigma 3 or 6) in training and testing (ablation) | Ours · synthetic LF, task 2 to 25 | PSNR: sigma 3, 6 | No clear effect |
Closest papers by meaning
- Deep Spatial-Angular Regularization for Light Field Imaging, Denoising, and Super-Resolution (TPAMI 2021 · similarity 0.98)
- Fast Light Field Reconstruction With Deep Coarse-To-Fine Modeling of Spatial-Angular Clues (ECCV 2018 · similarity 0.92)
- Learning to Capture Light Fields through a Coded Aperture Camera (ECCV 2018 · similarity 0.91)
- Probabilistic-Based Feature Embedding of 4-D Light Fields for Compressive Imaging and Denoising (IJCV 2024 · similarity 0.89)
- Light Field Super-Resolution using a Low-Rank Prior and Deep Convolutional Neural Networks (TPAMI 2019 · similarity 0.89)
- Time-Efficient Light-Field Acquisition Using Coded Aperture and Events (CVPR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
