Research map / Image generation / Super-resolution
ISP2HRNet: Learning to Reconstruct High Resolution Image from Irregularly Sampled Pixels via Hierarchical Gradient Learning
Yuanlin Wang, Ruiqin Xiong, Rui Zhao, Jin Wang, Xiaopeng Fan, Tiejun Huang
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 |
|---|---|---|---|
| replace RBF interpolation preprocessing with the full GD+GSE+IRC pipeline (ablation) | ISP2HRNet · Urban100, ratio 0.2 | PSNR: 22.60 -> 27.43 | Improved |
| add gradient derivation (GD) before IRC (ablation) | ISP2HRNet · Urban100, ratio 0.2 | PSNR: 26.11 -> 26.49 | Improved |
| add local gradient structure encoder (GSE) (ablation) | ISP2HRNet · Urban100, ratio 0.2 | PSNR: 26.49 -> 27.43 | Improved |
| use content-dependent IRC instead of distance-based Gaussian weighting (ablation) | ISP2HRNet · Urban100, ratio 0.2 | PSNR: 26.25 -> 27.43 | Improved |
| use dual-branch weight generator instead of single-branch (ablation) | IRC module · Urban100, ratio 0.2 | PSNR: 26.92 -> 27.43 | Improved |
| use dual-branch weight generator instead of single-branch (ablation) | IRC module · Urban100, ratio 0.4 | PSNR: 30.94 -> 31.40 | Improved |
| ISP2HRNet versus inpainting and restoration baselines | random missing pixel reconstruction · Urban100, ratio 0.2 | PSNR: 26.34 -> 27.43 | Improved |
| ISP2HRNet versus DPIR plus super-resolution pipelines at x2 and x4 on Urban100 | SR from incomplete image · Urban100, x2 and x4 | PSNR: 25.34 vs 25.69 (x2); 21.53 vs 21.74 (x4) | Hurt |
| apply model to sampling ratios outside training distribution (x3, x4) | ISP2HRNet · SR from incomplete image, DIV2K Val x4 | PSNR: 25.76 vs 25.74 (DPIR+SwinIR) | Mixed |
Closest papers by meaning
- D2IM-Net: Learning Detail Disentangled Implicit Fields From Single Images (CVPR 2021 · similarity 0.85)
- SkeletonNet: A Topology-Preserving Solution for Learning Mesh Reconstruction of Object Surfaces From RGB Images (TPAMI 2021 · similarity 0.84)
- Decompositional Neural Scene Reconstruction with Generative Diffusion Prior (CVPR 2025 · similarity 0.83)
- HUNet: Homotopy Unfolding Network for Image Compressive Sensing (CVPR 2025 · similarity 0.83)
- DEALing with Image Reconstruction: Deep Attentive Least Squares (ICML 2025 · similarity 0.83)
- Image Reconstruction Via Autoencoding Sequential Deep Image Prior (NeurIPS 2024 · similarity 0.83)
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.
