Research map / Image generation / Super-resolution
OmniSSR: Zero-shot Omnidirectional Image Super-Resolution using Stable Diffusion Model
Runyi Li, Xuhan Sheng, Weiqi Li, Jian Zhang
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 |
|---|---|---|---|
| add Gradient Decomposition correction on top of TP input (ablation) | OmniSSR (StableSR backbone) · ODI-SR x2 | FID: 43.91 -> 15.41 | Improved |
| add OTII without GD correction (ablation) | OmniSSR (StableSR backbone) · ODI-SR x2 | FID: 43.91 -> 65.35 | Hurt |
| add OTII together with GD correction at every step (ablation) | OmniSSR · ODI-SR x2 | WS-PSNR: 26.77 -> 28.58 | Improved |
| transform ERP input to TP images before StableSR (ablation) | StableSR · ODI-SR x2 | WS-PSNR: 22.69 -> 23.53 | Improved |
| add 4x pre-upsampling before ERP-TP and TP-ERP transforms (ablation) | OTII · ERP-TP-ERP round trip | WS-PSNR: 28.98 -> 38.18 | Improved |
| apply pre-upsampling only to the ERP before ERP-TP (4,1) (ablation) | OTII · ERP-TP-ERP round trip | WS-PSNR: 28.98 -> 28.99 | No clear effect |
| apply proposed OTII and GD techniques to a SwinIR backbone (ablation) | SwinIR · ODI-SR x4 | FID: 27.11 -> 13.33 | Improved |
| increase gamma_l in GD correction from 0 to 0.5 (ablation) | OmniSSR · ODI-SR x4 | WS-PSNR: 24.33 -> 25.77 (FID 27.05 -> 30.97) | Mixed |
| increase gamma_p from 0 to 1 in GD correction (ablation) | OmniSSR · ODI-SR x4 | FID: 39.64 -> 30.97 | Improved |
| use OmniSSR instead of supervised OSRT on ODI-SR x4 | OmniSSR vs OSRT · ODI-SR x4 | WS-PSNR: 26.89 -> 25.77 | Hurt |
| use OmniSSR instead of OSRT on ODI-SR x4 perceptual metrics | OmniSSR vs OSRT · ODI-SR x4 | LPIPS: 0.3258 -> 0.2977 | Improved |
| run ERP-TP transforms in the latent space rather than image space | OmniSSR · qualitative | image quality | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- OSRT — OSRT: Omnidirectional Image Super-Resolution With Distortion-Aware Transformer (CVPR 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Fast Omni-Directional Image Super-Resolution: Adapting the Implicit Image Function with Pixel and Semantic-Wise Spherical Geometric Priors (AAAI 2025 · similarity 0.88)
- Arbitrary-Scale Image Generation and Upsampling using Latent Diffusion Model and Implicit Neural Decoder (CVPR 2024 · similarity 0.88)
- Spherical Pseudo-Cylindrical Representation for Omnidirectional Image Super-resolution (AAAI 2024 · similarity 0.88)
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (CVPR 2024 · similarity 0.87)
- TSD-SR: One-Step Diffusion with Target Score Distillation for Real-World Image Super-Resolution (CVPR 2025 · similarity 0.87)
- Edge-SD-SR: Low Latency and Parameter Efficient On-device Super-Resolution with Stable Diffusion via Bidirectional Conditioning (CVPR 2025 · similarity 0.87)
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.
