Research map / Image generation / Super-resolution
LAR-SR: A Local Autoregressive Model for Image Super-Resolution
Baisong Guo, Xiaoyun Zhang, Haoning Wu, Yu Wang, Ya Zhang, Yan-Feng Wang
Check 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 local autoregressive texture generation with a learned texture codebook | LAR-SR · DIV2K validation ×4 | LPIPS: 0.121 -> 0.114 | Improved |
| Use local autoregressive texture generation with a learned texture codebook | LAR-SR · DIV2K validation ×4 | LPIPS: 0.110 -> 0.114 | Hurt |
| Use local autoregressive texture generation with a learned texture codebook | LAR-SR · CelebA test ×8 | LPIPS: 0.085 -> 0.077 | Improved |
| Use local autoregressive texture generation with a learned texture codebook | LAR-SR · FFHQ test ×8 | LPIPS: 0.100 -> 0.088 | Improved |
| Replace LAR layers with local mask attention (ablation) | LAR-SR · DIV2K validation ×4 | Time: 14.7s -> 7.8s | Improved |
| Replace LAR layers with local mask attention (ablation) | LAR-SR · DIV2K validation ×4 | LPIPS: 0.114 -> 0.118 | Hurt |
| Set patch size to 1×1 rather than the 4×4 configuration (ablation) | LAR-SR · DIV2K validation ×4 | LPIPS: 0.114 -> 0.163 | Hurt |
| Set patch size to 2×2 rather than the 4×4 configuration (ablation) | LAR-SR · DIV2K validation ×4 | LPIPS: 0.114 -> 0.120 | Hurt |
| Set patch size to 3×3 rather than the 4×4 configuration (ablation) | LAR-SR · DIV2K validation ×4 | LPIPS: 0.114 -> 0.116 | Hurt |
| Add a coarse-SR structural branch in Stage 1 and condition Stage 2 on coarse SR (ablation) | LAR-SR · DIV2K validation ×4 | LPIPS: 0.185 -> 0.114 | Improved |
| Add the coarse-SR structural input to the Stage 1 decoder (ablation) | LAR-SR · DIV2K validation ×4 | LPIPS: 0.164 -> 0.114 | Improved |
| Use LAR-SR instead of VQVAE+pixelCNN | CelebA ×8 · CelebA ×8 | Time: 230.7s -> 1.2s | Improved |
Later papers that compared with this method
1 later accepted paper reports LAR-SR as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Local Implicit Normalizing Flow for Arbitrary-Scale Image Super-Resolution (CVPR 2023 · better on 1 of 1 shared result)
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.
- SRFlow — SRFlow: Learning the Super-Resolution Space with Normalizing Flow (ECCV 2020) · better on 1 of 1 shared result
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
- Noise-Started One-Step Image Super-Resolution via LR-Conditioned SplitMeanFlow and GAN Refinement (NeurIPS 2026 · similarity 0.90)
- Visual Autoregressive Modeling for Image Super-Resolution (ICML 2025 · similarity 0.90)
- NSARM: Next-Scale Autoregressive Modeling for Robust Real-World Image Super-Resolution (NeurIPS 2026 · similarity 0.90)
- GSDD: Generative Space Dataset Distillation for Image Super-resolution (AAAI 2024 · similarity 0.89)
- Training Generative Image Super-Resolution Models by Wavelet-Domain Losses Enables Better Control of Artifacts (CVPR 2024 · similarity 0.88)
- Iterative Token Evaluation and Refinement for Real-World Super-resolution (AAAI 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.
