Research map / Image generation / Image restoration
Burst Image Deblurring Using Permutation Invariant Convolutional Neural Networks
Miika Aittala, Fredo Durand
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 max-pooling permutation-invariant fusion instead of recurrent frame-by-frame update | burst deblurring network · Delbracio et al. burst dataset example (c) | SSIM: 0.8764 -> 0.9578 | Improved |
| use permutation-invariant network instead of Fourier burst accumulation (FBA) | burst deblurring network · Delbracio et al. burst dataset example (c) | SSIM: 0.9456 -> 0.9578 | Improved |
| replace max-pooling with mean-pooling in the permutation invariant layers (ablation) | burst deblurring network · burst deblurring | restoration quality | No clear effect |
| add correlated realistic noise model to training instead of independent Gaussian noise with std 0.02 (ablation) | burst deblurring network · Figure 1 burst | noise level of output | Improved |
| add dynamic range expansion to training data (ablation) | burst deblurring network · extremely degraded burst | intensity of dark regions and streak handling | Improved |
| use entire burst instead of deblurring only the sharpest frame with single-image methods | burst deblurring network · self-captured noisy bursts | visual deblurring quality | Improved |
| train on single-image bursts only versus dedicated single-image deblurring methods | burst deblurring network · noisy vs less noisy single images | visual quality | Mixed |
Closest papers by meaning
- Iterative Residual CNNs for Burst Photography Applications (CVPR 2019 · similarity 0.90)
- Efficient Dynamic Scene Deblurring Using Spatially Variant Deconvolution Network With Optical Flow Guided Training (CVPR 2020 · similarity 0.88)
- Deep Image Deblurring: A Survey (IJCV 2022 · similarity 0.88)
- Burst Denoising With Kernel Prediction Networks (CVPR 2018 · similarity 0.88)
- Burst Image Restoration and Enhancement (CVPR 2022 · similarity 0.88)
- Region-Adaptive Dense Network for Efficient Motion Deblurring (AAAI 2020 · 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.
